The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several git commands to interact with the target repository.
Evidence: git rev-parse --show-toplevel, git rev-parse HEAD, git cat-file -e <sha>^{commit}, git diff --name-status <sha>..HEAD, git ls-files | wc -l, and git status --porcelain in SKILL.md.
Intent: These commands are used solely for metadata extraction and repository state tracking, which is consistent with the skill's described purpose.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) due to the ingestion of untrusted repository content.
Ingestion points: The skill reads files from the target repository in Step 2 (Cold path) and Step 4 (Patch) to build a memory map.
Boundary markers: None identified; the skill does not explicitly instruct the agent to ignore instructions embedded within the codebase files it reads.
Capability inventory: The skill has the capability to read any file in the repository and write output to its local memory/ directory using standard file operations.
Sanitization: No sanitization or escaping mechanisms are described for the content read from the target repository before it is processed by the agent.

recon