The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local bash scripts provided with the package (setup-templates.sh and update-progress.sh) to manage the workflow's state and file structure.
[COMMAND_EXECUTION]: During the implementation and validation steps, the skill runs project-specific commands such as pnpm run typecheck, pnpm run lint, and pnpm run test to verify code quality.
[COMMAND_EXECUTION]: The skill automates git operations including branch creation and management, and uses the GitHub CLI (gh) to create pull requests as part of its finalization process.
[PROMPT_INJECTION]: The skill uses internal 'Mandatory Execution Rules' and 'Protocols' to guide the agent through its systematic workflow. These are operational constraints rather than malicious instructions.
[PROMPT_INJECTION]: An indirect prompt injection surface exists during the analysis and review phases where the skill ingests content from the local codebase.
Ingestion points: Codebase files and library documentation read during the 'Analyze' and 'Examine' stages.
Boundary markers: The skill separates findings by appending them to specific, pre-created markdown files (e.g., 01-analyze.md, 05-examine.md).
Capability inventory: The skill can execute shell commands (pnpm, git, gh) and has full read/write access to the local project directory.
Sanitization: Ingested code content is processed as text for analysis, utilizing the underlying agent's standard safety guardrails.

apex