The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill utilizes local shell scripts (scripts/collect_repo_facts.sh and scripts/check_bilingual_readme.sh) to inspect the project structure. These scripts use standard tools such as find, rg, tree, and git, and follow best practices like variable quoting and pipefail settings.
[PROMPT_INJECTION] (LOW): The skill possesses an Indirect Prompt Injection surface (Category 8) because it reads and processes untrusted repository files to generate documentation. 1. Ingestion points: collect_repo_facts.sh output and direct reading of source/config files as per the workflow. 2. Boundary markers: No explicit delimiters or instructions are provided to the model to ignore embedded instructions in the source data. 3. Capability inventory: Ability to write to the local filesystem (README.md) and execute skill-provided scripts. 4. Sanitization: No sanitization of ingested content is performed before processing.
[DATA_EXPOSURE] (SAFE): While the skill reads project source code, it explicitly instructs the model not to call external APIs or README generation providers, ensuring data remains within the local session context.

readme-maintainer