The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user input and persists it into instruction files that influence future agent behavior.
Ingestion points: Processes user statements expressing rules, preferences, or corrections as defined in the SKILL.md activation triggers.
Boundary markers: The skill uses markdown headers and blockquotes to delimit captured content within .agents/memory/captured-rules.md.
Capability inventory: The skill possesses the capability to modify repository instruction files such as CLAUDE.md, AGENTS.md, and project memory files as described in the Capture Process of SKILL.md.
Sanitization: The skill includes explicit instructions for redacting secrets (API keys, tokens, passwords) and requires user confirmation before promoting temporary captures to permanent storage.

rules-capture