The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes project-specific shell scripts, such as scripts/pr-coordinator.sh and scripts/beads-context.sh, using arguments derived from the local environment and issue tracking data. These scripts represent unverified code execution within the project context.
[DYNAMIC_EXECUTION]: The skill relies on the execution of external shell scripts located in the project's directory which are not part of the skill's defined logic, constituting a dynamic execution surface.
[INDIRECT_PROMPT_INJECTION]: The skill reads pull request templates from the repository (e.g., .github/pull_request_template.md) and populates them with data for use in the gh pr create command. A malicious repository could use these templates to inject unwanted content or influence the agent's behavior.
Ingestion points: PR template files and project-local helper scripts (SKILL.md).
Boundary markers: The instructions do not include specific delimiters or 'ignore' instructions for the content read from templates.
Capability inventory: Shell execution (bash), Git operations, GitHub CLI (gh), and Forge CLI operations (SKILL.md).
Sanitization: No sanitization or validation processes are described for the template content or script outputs before they are processed or used in further commands.

ship