The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The documentation recommended installation method involves piping a remote script to a shell (curl | bash). This is a high-risk pattern, though standard for many CLI developer tools. The repository provides SHA-256 checksums to mitigate tampering risks.
[EXTERNAL_DOWNLOADS]: The install.sh and install.ps1 scripts download release packages from GitHub and verify them against published SHA-256 hashes. It also makes queries to api.osv.dev for dependency vulnerability data.
[COMMAND_EXECUTION]: The scripts/secrets.py utility uses subprocess.run to execute git log when performing repository history scans. scripts/scan.py and scripts/taint.py utilize the ast module to perform static analysis of Python code without executing it.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted source code. This creates a surface for indirect prompt injection where a malicious codebase could contain instructions designed to influence the AI's review behavior. The skill includes a false-positive-filter instruction set to mitigate this risk.
[CREDENTIALS_UNSAFE]: The repository includes numerous files containing hardcoded API keys, private keys, and connection strings (e.g., in tests/fixtures/ and tests/benchmark/secrets/). These are explicitly identified and used only as test fixtures for the scanner's detection engine.

code-vulnscan