The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes implementation plans (markdown files) as a source of truth for validation steps. A malicious plan could contain instructions that trick the agent into executing unintended commands during the validation phase.
Ingestion points: SKILL.md Step 1: "Read the implementation plan completely" and Step 2: "Systematic Validation".
Boundary markers: Absent; there are no instructions to treat the plan content as untrusted data rather than a set of instructions.
Capability inventory: Access to Bash and Task tools, with instructions to run make targets and git commands based on plan content.
Sanitization: Absent; plan content directly influences the agent's research tasks and verification steps.
[COMMAND_EXECUTION]: The skill invokes shell commands to perform its primary function of code validation.
Evidence: Execution of git log, git diff, and make targets (check, test, build, lint) defined in the project's build system via the Bash tool.

validate-plan