The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to build the polyscan tool (make -C polyscan build), run analysis (polyscan/polyscan analyze), manage repository downloads (curl, tar, rm), and interact with GitHub via the gh CLI. These operations are core to the skill's purpose as an automated auditing tool.
[EXTERNAL_DOWNLOADS]: The skill downloads source code tarballs from codeload.github.com and commit information from the GitHub API. These downloads target a well-known service and are necessary for the skill's function of auditing external repositories.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code from external repositories which is then passed to sub-agents for triage and classification. This creates a surface where malicious instructions embedded in the analyzed code (e.g., in comments or string literals) could attempt to influence the sub-agent's judgment or the resulting automated issue filing.
Ingestion points: Source code from external repositories (JS, TS, Go, Rust, C++) is read and sampled in SKILL.md (Step 5).
Boundary markers: The triage sub-agent is instructed to perform "Read-only triage" and "Do not modify any files," though it lacks explicit warnings to ignore instructions embedded within the analyzed code snippets.
Capability inventory: The skill can execute shell commands, perform network operations via curl and gh, and file public GitHub issues on the ludo-technologies/polyscan repository.
Sanitization: The skill uses jq for structured data extraction and applies rate limits and deduplication logic before auto-filing issues.

polyscan-fp-audit