The Agent Skills Directory

[SAFE]: No malicious patterns, unauthorized data access, or exfiltration vectors were identified. The skill's operations are confined to local workspace validation and adhere to its stated purpose as an advisory quality gate.
[COMMAND_EXECUTION]: The skill executes local shell commands (git, grep, awk, test, wc) to perform rule-based validation of the plan.md file. These commands are used strictly for structural and presence checks (e.g., verifying headers, timestamps, and file paths) within the repository context.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted data from plan.md to trigger scoring and automated fixes.
Ingestion points: Content from plan.md is read via grep and awk and provided to the LLM for dimensional scoring.
Boundary markers: No explicit instruction-isolation delimiters or boundary markers are defined for the content of the plan.
Capability inventory: The skill uses read capabilities (shell text processing) and write capabilities (automatic fixes in 'Fix Mode').
Sanitization: No specific sanitization or escaping is performed on the content read from plan.md before it is processed by the agent.

confidence