The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform structural filesystem operations including moving files (mv), creating directories (mkdir), and reverting changes via git checkout.
[COMMAND_EXECUTION]: Automated post-tool hooks execute a local Python script (validate_kb.py) using the uv run command to verify the integrity of the knowledge base after modifications.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes the content of markdown files to make decisions about terminology unification and file splitting.
Ingestion points: The compacter agent recursively reads .md files in the target directory (agents/compacter.md).
Boundary markers: There are no explicit delimiters or markers used to distinguish between data content and instructions when reading file contents.
Capability inventory: The agent has access to Bash, Write, and Edit tools which could be exploited if malicious instructions are followed.
Sanitization: The skill does not implement sanitization or filtering for the content ingested from the markdown files.

kb-compact