The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, specifically user-provided briefs and referenced project files such as documentation, Architecture Decision Records (ADRs), and invariant lists.
Ingestion points: references/homework.md instructs the agent to read the initial brief and all files it references to establish ground truth.
Capability inventory: The agent possesses the ability to execute shell commands for environment probes, run verifier scripts, and manage GitHub issues using the gh CLI tool.
Boundary markers: The instructions do not explicitly mandate the use of delimiters or 'ignore embedded instructions' warnings when processing external files.
Sanitization: No explicit sanitization or escaping mechanisms are defined for content extracted from external sources before it is processed in the critique or interview phases.
[COMMAND_EXECUTION]: The skill utilizes the GitHub (gh) command-line tool to interact with the environment's repository and issue tracker.
Evidence: references/spec-output.md uses gh repo view to verify the repository context and gh issue create/edit to publish the resulting technical specification as a GitHub issue.
[DYNAMIC_EXECUTION]: The workflow requires the agent to identify, generate, and execute scripts to verify current behavior and proposed implementation goals.
Evidence: SKILL.md and references/homework.md define 'probes' and 'verifiers' (which can be scripts or scripted user flows) to validate technical claims. The agent is instructed to run these verifiers within the project's verification environment (staging, test accounts, etc.).

idea-to-spec