The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to read .git/config and execute git remote -v to identify the repository host and URL. While this accesses metadata about the user's environment, it is necessary for the skill's stated purpose of configuring the issue tracker.
[COMMAND_EXECUTION]: The templates provided for GitHub and GitLab issue tracking (issue-tracker-github.md, issue-tracker-gitlab.md) contain instructions for the agent to execute specific CLI commands using gh and glab, including API calls (gh api, glab api) to manage issue dependencies and sub-issues.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8). It reads multiple repository files (e.g., WORK-ROUTING.md, ARCHITECTURE.md, CONTEXT.md) and uses their content to determine how to scaffold new documentation.
Ingestion points: Reads WORK-ROUTING.md, .git/config, ARCHITECTURE.md, PLANS.md, and other documentation files from the target repository root and docs/ directory.
Boundary markers: The instructions lack explicit delimiters when reading these files, though they do instruct the agent to treat existing documents as "repository doctrine."
Capability inventory: The agent can read/write files and execute gh/glab CLI commands.
Sanitization: The skill includes a "Confirm and edit" step (Step 3) where the agent must show the user a draft of the changes before writing, which serves as a human-in-the-loop security control.

setup-matt-pocock-skills