The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes instructions to execute a bundled Python script, scripts/check_workspace_fixtures.py. This script performs structural integrity checks on the local workspace by comparing files against defined templates. It uses standard Python libraries and does not perform network operations or access system paths outside of the workspace.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses local workspace metadata (workspace.toml) and markdown content within the workspace directory to generate indexes and statistics. All operations are local to the user's filesystem, and no data exfiltration to external domains was detected. The use of .env files for secret management is documented as a safe practice.\n- [INDIRECT_PROMPT_INJECTION]: The skill aggregates content from multiple local wikis, creating a surface for indirect prompt injection if the source data contains malicious instructions. This is mitigated by the skill's use of structured boundary definitions and its reliance on predefined templates. Ingestion points: scan and query workflows reading local markdown files in wiki subdirectories. Boundary markers: explicit scope declarations in AGENTS.md and CLAUDE.md. Capability inventory: local file operations and bundled script execution. Sanitization: use of structured TOML metadata and markdown templates for content aggregation.

yzr-llm-workspace-management