The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted specification documents (such as PRDs, SRS, or User Stories) during the INGEST phase, which creates a potential surface for indirect prompt injection attacks.
Ingestion points: External specification documents provided as input to the attest command in SKILL.md.
Boundary markers: The agent is instructed to quote acceptance criteria (AC) verbatim and apply AMBIGUOUS_FLAG markers to any subjective or incomplete content, as detailed in reference/criteria-extraction.md.
Capability inventory: The skill's operations are limited to static analysis, documentation generation, and report aggregation. It is explicitly prohibited from modifying implementation code or performing runtime execution.
Sanitization: Employs an ISO/IEC/IEEE 29148 quality gate to validate that requirements are necessary, verifiable, and unambiguous before they are extracted as criteria.
[EXTERNAL_DOWNLOADS]: The skill contains references to several well-known and reputable third-party development and testing services.
Evidence: Documentation links to schemathesis.io, tracetest.io, pactflow.io, and reqnroll.net are included in reference/modern-tooling.md.
Context: These references are provided as informational recommendations to guide users toward modern verification tooling and do not represent automated installation or execution of untrusted scripts.

attest