The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted content from external repositories, including source code, documentation, and metadata files which may contain malicious instructions.
Ingestion points: The skill reads AGENTS.md, repository history, documentation, and source code as specified in SKILL.md.
Boundary markers: The instructions lack specific delimiters or "ignore embedded instructions" warnings for the agent when processing these external files.
Capability inventory: The skill enables shell execution for "minimal reproductions," tests, and "controlled measurements."
Sanitization: No explicit sanitization or filtering of external content is required before interpolation into the agent's context.
[DYNAMIC_EXECUTION]: The skill instructs the agent to generate and run code snippets or "fixtures" to verify claims.
Evidence: SKILL.md describes running "generated code and state" and creating "minimal reproductions" within "disposable workspaces."
[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands and test suites in the local environment to gather evidence.
Evidence: Both SKILL.md and references/test-effectiveness.md involve executing "test entry points," "command effects," and "actual UI use" to settle factual uncertainties.

investigate-with-evidence