The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill retrieves source files (github.ts and github-registration.test.ts) from a project branch (channels) using the nc:copy directive and writes them to the local src/channels/ directory.
[COMMAND_EXECUTION]: Several shell commands are executed, including pnpm run build for compilation, vitest for unit testing, and service restart commands (launchctl or systemctl). These are used to validate the installation and apply changes.
[EXTERNAL_DOWNLOADS]: The skill installs the @chat-adapter/github package at version 4.29.0 using pnpm. This is a standard dependency installation from a known registry.
[PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by enabling the agent to process GitHub issue and pull request comments.
Ingestion points: External comments provided via GitHub webhooks are processed by the src/channels/github.ts adapter.
Boundary markers: The provided instructions do not specify the use of delimiters or 'ignore' instructions for the ingested content.
Capability inventory: The agent is granted repository write access via a fine-grained Personal Access Token to post responses.
Sanitization: No explicit sanitization or filtering logic is detailed in the setup instructions.
[CREDENTIALS_UNSAFE]: The skill provides instructions for capturing a GitHub Personal Access Token and a webhook secret and storing them in a local .env file. This follows secure development practices by keeping secrets out of the codebase while ensuring they are available to the runtime environment.

add-github