The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several standard shell utilities including git, ls, grep, sed, and comm to perform gap analysis on the repository's history and detect missing daily log files. These operations are constrained to the local environment and the current project's context.
[COMMAND_EXECUTION]: Provides a cleanup mechanism for stale experiment folders using rm -rf. This operation is limited to subdirectories within the experiments/ path and requires explicit verbal confirmation from the user before execution.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from git commit messages to reconstruct historical daily logs and propose new project rules.
Ingestion points: Commit messages retrieved via git log and pattern entries from .claude/memory/MEMORY.md.
Boundary markers: Reconstructed (backfilled) logs are explicitly prefixed with a header identifying them as lower-detail entries derived from git history.
Capability inventory: The skill has the ability to execute shell commands (git, rm), write to local markdown files (daily/, .claude/rules/, knowledge/concepts/), and read from the project's knowledge base.
Sanitization: No explicit content sanitization is mentioned for the commit messages before they are written to project documentation, though the risk is minimal given the developer-centric use case.

close-day