The Agent Skills Directory

[DATA_EXFILTRATION]: The skill reads .git/config and runs git remote -v to identify the repository's origin. While standard for repository analysis, the .git/config file is a sensitive path that may contain credentials, access tokens, or private repository URLs.
[COMMAND_EXECUTION]: The skill executes git remote -v to gather information about the repository environment. Additionally, it provides templates that instruct the agent to use the gh and glab CLI tools for issue management.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its ingestion of untrusted repository data.
Ingestion points: Reads content from AGENTS.md, CLAUDE.md, CONTEXT.md, CONTEXT-MAP.md, and the docs/adr/ directory to understand the project state.
Boundary markers: None identified in the prompt instructions to distinguish between the skill's instructions and the content of the files being read.
Capability inventory: The skill has file-write capabilities, modifying AGENTS.md or CLAUDE.md and creating several files in the docs/agents/ directory.
Sanitization: No explicit sanitization or validation of the ingested file content is performed before it is presented to the user or used to generate new documentation.

setup-matt-pocock-skills