The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill architecture involves processing untrusted external inputs, such as user stories and PR file changes, which creates a potential surface for indirect prompt injection.
Ingestion points: The skill ingests userStories for requirements validation and prFiles for regression risk analysis as shown in the agent coordination snippets in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat processed data as passive content or to ignore potential instructions embedded within the requirements or code.
Capability inventory: The skill possesses capabilities to generate code/tests via qe-test-generator and interact with CI/CD environments through npm scripts.
Sanitization: No sanitization or filtering logic is described for the ingested text before it is utilized by the AI agents.
[EXTERNAL_DOWNLOADS]: The skill documentation includes examples of executing tools from external package registries.
Evidence: SKILL.md contains a GitHub Actions example that executes npx coverage-check --min 80. This command downloads and runs the coverage-check package from the npm registry. As this targets a well-known service, it is considered a safe and standard development practice.

shift-left-testing