The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted TDD artifacts which could contain malicious instructions designed to manipulate the audit outcome or score.
Ingestion points: The skill reads files from paths matching docs/07_TDD/TDD-NN_*/... to perform structural and content reviews.
Boundary markers: The skill employs explicit instructions for lenses to 'Disregard the author self-assessment score' and citing specific playbook checks to prevent the model from being anchored by claims within the untrusted document.
Capability inventory: The skill writes audit reports and saga journals to the .aidoc/ directory, creates review directories, and dispatches various specialized subagents (e.g., security-engineer, chaos-engineer).
Sanitization: No programmatic sanitization of the artifact content is performed; the skill relies on instructional de-anchoring and subagent briefs.
[COMMAND_EXECUTION]: The skill uses bash commands for state management and timing, such as mkdir -p for directory creation and date +%s for calculating the elapsed time against the 1500s SOFT_DEADLINE. These commands operate on local paths and use internal artifact identifiers.

doc-tdd-audit