The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates for creating implementation plans that include shell commands for running tests (e.g., pytest) and performing version control operations (e.g., git commit). These commands are standard developer utilities and are used appropriately within the context of the skill's intended purpose for software implementation.
[INDIRECT_PROMPT_INJECTION]: The skill is intended to process external specifications and requirements documents to generate structured plans. This workflow constitutes a potential attack surface for indirect prompt injection if the input specifications contain malicious instructions designed to influence the agent's behavior during plan generation.
Ingestion points: The agent processes external data from a "spec" or "design doc" as referenced in the SKILL.md file.
Boundary markers: The skill uses a structured plan document header and specific task boundaries to keep context focused, though it does not explicitly use escaping for the spec content.
Capability inventory: The generated plans involve file modification and the execution of shell commands for testing and source control management (SKILL.md).
Sanitization: The skill mitigates injection risks by enforcing a mandatory "Self-Review" checklist and utilizing a separate "Plan Document Reviewer" subagent (plan-document-reviewer-prompt.md) to verify that the plan accurately reflects the specification without introducing unrequested changes.

writing-plans