The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates indirect prompt injection by instructing the agent to ingest and process arbitrary repository data (source code, manifests, and documentation) to generate documentation deliverables.
Ingestion points: Reads content from README files, manifests (e.g., package.json), CI configurations, source code entrypoints, and existing documentation assets (references/evidence-research.md, references/agents-md.md).
Boundary markers: The instructions do not provide specific delimiters or guidance for isolating untrusted repository data from the agent's internal instruction context.
Capability inventory: The skill allows for writing or overwriting files in the local environment, which is controlled by a user-approval gate (SKILL.md).
Sanitization: No automated sanitization or filtering logic is defined for processed input; the skill relies entirely on manual verification and post-generation checklists (references/write-verify.md).

octocode-documentation