The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions to execute shell commands using sed and grep to find and read project-specific memory files located in $HOME/.claude/projects/.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests data from external sources (user descriptions and project memory) and interpolates them into generated files without automated sanitization or strict boundary markers.
Ingestion points: Step 1 defines ingestion from auto-memory files in $HOME/.claude/ and user-supplied descriptions.
Boundary markers: The interpolation template in Step 5 does not utilize delimiters or warnings to ignore potentially malicious embedded instructions in the source text.
Capability inventory: The skill spawns a skill-extractor agent that has the capability to create new directories and write multiple files to the filesystem.
Sanitization: The skill relies on manual quality gates (Step 6) rather than programmatic validation or escaping of the interpolated content.

extract