The Agent Skills Directory

[COMMAND_EXECUTION]: The scan.mjs script contains a command injection vulnerability. The --since command-line argument is extracted and directly interpolated into a shell command string passed to execSync without any validation or escaping. A malicious string provided as an argument could execute arbitrary shell commands in the environment where the skill is run.
Evidence: The interpolation of the ${SINCE} variable into the git diff command string within scan.mjs.
[REMOTE_CODE_EXECUTION]: The script scan.mjs executes shell commands via execSync using user-controllable input from CLI arguments, which constitutes a code execution risk if parameters are manipulated.
Evidence: execSync(git diff --name-only ${SINCE}...HEAD -- ..., ...) in scan.mjs.
[DYNAMIC_EXECUTION]: The release-gate.mjs script uses dynamic import() to load check modules from a computed file path, which is a form of dynamic code loading from local paths.
Evidence: const mod = await import(pathToFileURL(modPath).href); in release-gate.mjs.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository's source code and documentation, creating an attack surface where malicious content could influence the scanner's output.
Ingestion points: scan.mjs reads source code (.ts) and markdown (.md) files throughout the repository using fs.readFileSync and custom directory traversal.
Boundary markers: Absent. The script does not use delimiters or instructions to ignore potential injection content within the files it scans.
Capability inventory: execSync (in scan.mjs for executing git commands), fs.readFileSync (multiple files), fs.lstatSync (in checks/skill-internal-distribution.mjs).
Sanitization: Absent. The ingested content is processed using regular expressions without escaping or sanitization of the data before it is used to generate findings.

cli-self-check