The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it analyzes untrusted repository content (code, docs, and structure) to produce reports and harness artifacts. Malicious content within the repository could potentially influence the agent's logic. * Ingestion points: Processes repository-wide files during the Phase 1 assessment as defined in SKILL.md and references/assessment.md. * Boundary markers: Absent; no instructions are provided to the agent to distinguish between its own logic and instructions found within analyzed files. * Capability inventory: Generates documentation (AGENTS.md, docs/), configuration (.harness/*.yml), lint rules, CI workflows (GitHub Actions), and validation scripts as described in SKILL.md. * Sanitization: Absent; the skill does not specify any validation or sanitization of content extracted from the repository.
[COMMAND_EXECUTION]: The skill involves the creation of executable files and configurations to enforce repository standards. * Evidence: SKILL.md Phase 5 (Enforcement) instructs the agent to generate lint configurations, GitHub Actions workflows, structural tests, and validation scripts for repository maintenance.
[EXTERNAL_DOWNLOADS]: The skill metadata references external documentation from trusted and well-known sources. * Evidence: metadata.json includes links to official OpenAI resources and reputable technical blogs.

harness-engineering