The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from a session.json ledger and arbitrary files within a provided worktree path. (1) Ingestion points: External data is ingested via the session ledger (session.json) and files accessed through the --worktree argument. (2) Boundary markers: The instructions do not define specific delimiters or boundary markers to separate external content from the agent's primary instructions. (3) Capability inventory: The skill's capabilities are limited to reading local files and generating a report. It does not invoke network tools, write to the filesystem, or execute shell commands. (4) Sanitization: The skill contains a strict 'Never Quote' policy that forbids the reproduction of raw text or file contents in the output, providing a robust sanitization measure against data leakage.

session-retro