The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends downloading and installing several extensions from untrusted GitHub repositories in references/ecosystem-skills.md.
Instructions include npx skills add commands for repositories owned by individual users: nextlevelbuilder/ui-ux-pro-max-skill, leonxlnx/taste-skill, and emilkowalski/skills.
The skill also references official and trusted sources, including backstage.io and the anthropics/skills repository.
[COMMAND_EXECUTION]: The skill guides the agent to execute shell commands to analyze the environment and manage the build lifecycle.
Executes a bundled script: python scripts/detect-rhdh-context.py --path <plugin-dir> which performs file system analysis.
Reads workspace-specific instruction files using cat AGENTS.md and cat CLAUDE.md.
Utilizes standard toolchains for development including yarn build, yarn test, and yarn tsc.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by reading data from the local project environment.
Ingestion points: The script detect-rhdh-context.py recursively reads content from all .ts and .tsx files within the src/ directory of the target project. Additionally, the skill explicitly reads AGENTS.md and CLAUDE.md from the workspace root.
Boundary markers: No delimiters or explicit instructions to ignore embedded instructions are used when reading these files into the agent context.
Capability inventory: The agent has the capability to execute shell commands (via python and yarn), perform Git operations, and write to the filesystem (e.g., generating API reports and changesets).
Sanitization: No sanitization, filtering, or escaping is performed on the content of the ingested files before they are interpreted by the agent.

rhdh-coding