The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to interact with the local environment. Specifically, it uses git symbolic-ref, git remote show, git merge-base, git diff, and git log to analyze the repository state. Additionally, the 'Self-Healing Validation' phase involves executing 'generated tests' and checking for compilation or syntax errors, which implies the execution of generated code or scripts.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted external data from branch diffs, commit messages, and external specification files (stories, PRDs, specs).
Ingestion points: Data enters the context via git diff outputs, git log commit history, and the contents of provided specification files (e.g., from /spec or local paths).
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the processed external data.
Capability inventory: The agent has the capability to execute shell commands (git), run generated test code, and invoke other agent skills like /story-implementer and /evolve.
Sanitization: There is no mention of sanitizing or escaping the content of the diffs or specs before they are analyzed by the LLM.
[DATA_EXPOSURE]: The skill attempts to locate and write to a specific directory in the user's home folder (~/.claude/projects/) to record 'Self-Evolution Telemetry' in a file named skill-telemetry.md. While this is a form of local state tracking, it involves accessing hidden platform-specific configuration paths.

manual-test-plan