The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions and install hooks reference a setup script hosted on the author's repository (github.com/Gamaroff/agent-skills). This script is designed to be executed via bash process substitution to configure the local environment.
[INDIRECT_PROMPT_INJECTION]: The orchestrator ingests data from external tracking platforms such as GitHub and Jira (specifically issue and pull request bodies). This represents an attack surface for indirect prompt injection where malicious instructions could be embedded in ticket descriptions.
Ingestion points: references/develop-pipeline-step-0-resolve-and-prepare.md (via gh issue view).
Boundary markers: The skill uses structured subagent prompts that require responses in specific JSON or YAML formats, which acts as a delimiter.
Capability inventory: The agent has access to shell execution, file system modification, and network requests via CLI tools.
Sanitization: The skill uses scripts like github-body-text.js to escape control characters and report-lint.js to validate the structure of ingested and generated content.
[DYNAMIC_EXECUTION]: The skill includes a verification engine (references/qa-execute-snippets.mjs) that extracts and runs shell commands found in documentation to ensure they are functional. Security risks are mitigated by the implementation of a command allow-list and the use of sandboxed temporary directories for execution.

develop-story