The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external sources including requirement documents (PRD), API documentation, bug reports, and FAQ. This creates a surface for potential indirect prompt injection attacks.
Ingestion points: Requirement documents (PRD), API documentation, Bug reports, FAQ sections, and the output of git diff commands as specified in 'Stage 1' of the SKILL.md.
Boundary markers: The skill does not explicitly instruct the agent to use specific delimiters (e.g., XML tags or unique markers) or 'ignore embedded instructions' warnings when parsing the content of these external documents.
Capability inventory: The skill leverages an 'Explore' tool for recursive file reading, executes git shell commands for repository analysis, and runs a Python script for schema validation.
Sanitization: No evidence of sanitization, regex filtering, or escaping of the user-provided or document-provided content was found.
[COMMAND_EXECUTION]: To achieve its goal of code-prioritized testing, the skill explicitly instructs the agent to execute shell-level commands.
Evidence: The workflow includes running git diff <base>...<head> --stat to identify code changes and executing python ../core/scripts/validate_schema.py to validate generated YAML schemas.
[DYNAMIC_EXECUTION]: The skill employs an agentic workflow that involves spawning and managing multiple sub-agents.
Evidence: Stage 1, Step 3 describes the deployment of 1 to 3 'Explore' agents (proxies) focused on specific subsystems to perform concurrent code reviews following a predefined prompt framework.

test-case-writing