The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, including repository source code, configuration files, and external contracts supplied by users during an audit.
Ingestion points: Workflow step 2 in SKILL.md and document resolution in references/specification-authorities.md describe loading external artifacts into the agent's context.
Boundary markers: The skill defines a strict 'Rule Priority' hierarchy to resolve conflicts between user requests, repository guidance, and various contract types.
Capability inventory: The skill is intended to be read-only, producing findings and observations. It relies on file system access to read target code and can execute a local validation script.
Sanitization: The skill employs a formal JSON Schema (assets/frontend-visual-evidence.schema.json) and a custom Python validator (scripts/validate-frontend-visual-evidence.py) to enforce data integrity for visual evidence reports.
[COMMAND_EXECUTION]: The skill instructs the agent to execute specific commands as part of the audit workflow.
Evidence: SKILL.md mentions running git status --short for reproducibility checks, and references/frontend-visual-evidence.md specifies running python3 scripts/validate-frontend-visual-evidence.py to validate evidence artifacts. These commands are standard for a code auditing tool and target local repository state or provided utility scripts.

audit-frontend