The Agent Skills Directory

[COMMAND_EXECUTION]: Executes repository inspection commands. Evidence: The skill runs git status and git diff --stat to identify modified files and the scope of implementation changes.
[PROMPT_INJECTION]: Vulnerable to indirect prompt injection through repository content processing. 1. Ingestion points: Reads implementation plans, docs/architecture.md, docs/ledger.jsonl, and git diff output. 2. Boundary markers: No explicit delimiters or instructions are provided to the agent to disregard instructions found within the ingested files. 3. Capability inventory: Can read/write files in the local docs/ directory and execute git CLI tools; no network or arbitrary code execution capabilities were detected. 4. Sanitization: Content from files is summarized and archived without sanitization, though a human-in-the-loop confirmation step is required before writing changes.

evo-archiving