The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes source code and configuration files from the repository, creating an attack surface for instructions embedded in data.
Ingestion points: Reads files from src/lib/, package.json, and .claude/tend/config.yaml as per Phase 0 and 1 of SKILL.md.
Boundary markers: The skill does not provide instructions to ignore directives or prompts that might be embedded within the source code or configuration file comments.
Capability inventory: The skill executes shell commands to run repository-defined test runners and a local validation script, and it writes new content to test files.
Sanitization: While the skill uses a script to validate the structure of the output test file, it does not perform sanitization of the ingested input content.
[COMMAND_EXECUTION]: The skill identifies and executes the project's configured test runner and a local validation script.
Evidence: SKILL.md Phase 1 and 3 mandate running package.json test scripts and node scripts/validate-test-shape.mjs. Phase 2 and Reference 2/3 involve using grep commands to scan the repository.

tend-tests