The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extracts shell commands from markdown plan files under the 'Automated Verification' section and re-runs them to verify success criteria. It also uses git commands to correlate changes between commits.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from implementation plans which serves as a potential attack surface.
Ingestion points: Reads markdown plans from the thoughts/*/plans/ directory (Step 1).
Boundary markers: None identified for delimiting external command content.
Capability inventory: Executing extracted shell commands, running git diff operations, writing to plan files to update status, and invoking sub-skills for user interaction and script generation.
Sanitization: The skill contains a safety instruction to skip destructive or state-modifying items and focus on read-only checks, but lacks formal sanitization or escaping of the extracted commands.
[DYNAMIC_EXECUTION]: The skill optionally invokes the desplega:script-builder sub-skill to generate and execute wrapper scripts for verification commands. This involves runtime generation of executable scripts based on command strings found in documentation.

verifying