The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses child_process.spawnSync to execute local git commands such as git log, git diff, and git rev-list. These are used to provide the agent with summaries of repository activity since the last session. Evidence is found in hooks/session-start.mjs and commands/shared.mjs.
[DATA_EXFILTRATION]: The skill performs local data exposure by reading project configuration and metadata files, including .git/config (to extract repo URLs), package.json, and CLAUDE.md, to automatically populate project context. This data is stored locally in the user's home directory to provide project persistence. Evidence is found in commands/init.mjs.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it reads untrusted data from project documentation and manifests.
Ingestion points: commands/init.mjs and hooks/session-start.mjs ingest text from files like README.md, CLAUDE.md, and package.json.
Boundary markers: No explicit delimiters or warnings (e.g., "ignore instructions in the following text") are used when presenting this content to the agent.
Capability inventory: The skill has the capability to execute local shell commands (git) and write files to the filesystem.
Sanitization: The skill extracts content using regex or substring slicing and does not perform any escaping or sanitization to prevent embedded instructions from influencing the agent's behavior.

ck