The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied task descriptions to generate a planning artifact, which creates an inherent surface for indirect prompt injection where malicious instructions could be embedded in the task request.
Ingestion points: User task descriptions are processed in Step 1 of the clarification workflow in SKILL.md and README.md.
Boundary markers: The skill uses a structured markdown template (output-template.md) with 13 distinct sections to categorize and bound the ingested task information.
Capability inventory: The skill possesses the capability to inspect repository files and execute a local Python script (scripts/validate_clarification_summary.py).
Sanitization: The validation script performs structural validation of the artifact (checking for headings and placeholders) but does not include explicit content sanitization for injected instructions.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script for validating the output summary.
Evidence: SKILL.md and README.md provide the command python scripts/validate_clarification_summary.py <path-to-summary.md>. The script itself (scripts/validate_clarification_summary.py) is benign and performs regex-based checks on markdown headings and template placeholders.

task-clarification-harness