The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands and local scripts to perform its audit. The scripts/scout_repository.py script uses subprocess.run to call the GitHub CLI (gh) for fetching metadata. The dynamic workflow in scripts/repo-health-check.workflow.js instructs sub-agents to use find, grep, and gh to audit files, PRs, and issues. These operations are diagnostic in nature and use list-based arguments to avoid shell injection.
[INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection because it ingests untrusted data from the repository being audited, specifically PR titles, issue bodies, and marketplace manifest fields. However, this is mitigated by strict instructional guardrails in SKILL.md and references/health-check-methodology.md, which mandate that the agent must counter-verify all findings using direct commands (e.g., grep, sed, gh repo view) rather than relying on the sub-agents' summarized output.

marketplace-health-check