The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a 'hooks' architecture where it explicitly reads and follows instructions from project-specific files: PRE_PLAN.md and POST_PLAN.md. This design allows the agent's behavior to be directed by external data.
Ingestion points: The skill instructions in SKILL.md (Step 3 and Step 7) direct the agent to read and 'execute instructions' from configuration files located at <root>/config/hooks/PRE_PLAN.md and <root>/config/hooks/POST_PLAN.md.
Capability inventory: The skill has capabilities to read/write local files and execute bundled utility scripts. The hook instructions could potentially influence the agent to perform unintended file modifications or data processing.
Boundary markers: There are no specific boundary markers or safety instructions provided to the agent to distinguish between its core logic and the instructions found in the hook files.
Sanitization: No sanitization or schema validation is performed on the natural language instructions contained within the hook files.
[EXTERNAL_DOWNLOADS]: The scripts/check-for-updates.cjs script performs a network request to the GitHub API to check for newer versions of the tool.
The request targets https://api.github.com/repos/e0ipso/strikethroo/releases/latest. This is a vendor-owned resource matching the skill author ('e0ipso') and is used for version comparison only.
[COMMAND_EXECUTION]: The skill relies on the execution of several bundled CommonJS scripts (find-strikethroo-root.cjs, check-for-updates.cjs, and validate-plan-blueprint.cjs) to perform project discovery, version checks, and plan metadata validation.

st-refine-plan