The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides configuration instructions for the user to install the octocode-mcp tool using npx -y. While this is presented as a setup step for the user, it involves downloading and executing code from a public registry.
[COMMAND_EXECUTION]: During the verification phase (Phase 4), the agent is instructed to execute shell commands such as npm run build, npm run lint, and npm test to confirm the integrity of implementation changes.
[PROMPT_INJECTION]: The skill's architecture involves ingesting data from external research and local codebase analysis to generate implementation plans and code edits, creating a surface for indirect prompt injection.
Ingestion points: External repository findings and local codebase analysis delegated to octocode-researcher and octocode-research skills (SKILL.md).
Boundary markers: None explicitly defined to separate untrusted research data from the planning logic.
Capability inventory: Filesystem Write access and the ability to execute shell commands for building and testing code (SKILL.md).
Sanitization: Relies on a 'Decision Framework' with confidence levels and mandatory human approval gates before implementation phases.

octocode-plan