The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates its workflow by calling several local Python scripts via the shell. These include verify_required_files.py for stage gate validation, fact_check_gate.py for evidence verification, auto_clean_hook.py for final delivery processing, and update_run_manifest.py for state management. These are used to maintain the integrity of the multi-stage process.
[INDIRECT_PROMPT_INJECTION]: As the skill is designed to process external research materials and user-provided drafts, it identifies a potential surface for indirect prompt injection. It mitigates this risk with an explicit instruction to treat research findings as data and strictly avoid executing any instructions found within them. Furthermore, the mandatory multi-stage review process (Stages 7-9) and the final automated fact-checking gate (Stage 10.5) provide architectural defenses against malicious content propagation.

workflow-producer