The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from merged Pull Requests and source code, creating an indirect prompt injection surface.\n
Ingestion points: gh pr list output (PR titles, bodies, and file metadata) and project source code files (SKILL.md Step 1 and Step 2).\n
Boundary markers: Absent. The instructions do not specify any delimiters or warnings to ignore embedded instructions within the ingested PR data.\n
Capability inventory: File system writes to documentation (README, user guides), git commit/branch/PR operations, and repository survey via gh and git (SKILL.md Steps 4, 5, and 7).\n
Sanitization: Absent. While the skill performs classification and verification against source code, it does not implement sanitization or validation of the PR content before placing it into the documentation.\n- [COMMAND_EXECUTION]: Uses system commands such as git, gh, and grep to analyze the repository state, verify documentation, and perform commits. This includes managing multi-account environments using gh auth switch based on identifiers found in project-level configuration files (CLAUDE.md or AGENTS.md).

docs-from-prs