The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses npx octocode to run its core tool, which downloads and executes code from the npm registry at runtime.
[COMMAND_EXECUTION]: Instructions in references/code-check-loop.md and references/loop-protocol.md direct the agent to run 'Build / test run' to verify findings. This involves executing arbitrary shell commands within repositories that may be untrusted.
[EXTERNAL_DOWNLOADS]: The skill facilitates cloning remote repositories from GitHub and npm for analysis using tools like ghCloneRepo or octocode clone, as described in references/tools.md.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted data from GitHub, npm, and the web to guide its execution loop.
Ingestion points: SKILL.md and references/tools.md indicate data is read via ghSearchCode, npmSearch, and web surfaces.
Capability inventory: Command execution (build/test), file system access, and npx usage.
Boundary markers: No explicit delimiters or instructions are used to ignore embedded data prompts.
Sanitization: No sanitization of external repository content or search results is described.

octocode-loop