The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documentation files (e.g., markdown files, CHANGELOG, and reports) to identify edits and consolidation opportunities, creating a surface for indirect instructions to influence the agent.
Ingestion points: Reads content from files in docs/, book/, wiki/, and project root (CHANGELOG.md, README.md).
Boundary markers: Missing explicit delimiters or warnings to ignore embedded instructions in the processed data.
Capability inventory: Includes the ability to delete files (rm), modify project files via Agent(scribe:doc-editor), and execute powerful shell commands (jq, grep, awk).
Sanitization: While it employs style and 'slop' filters, these are focused on writing quality rather than neutralizing adversarial instructions.
[COMMAND_EXECUTION]: The workflow relies heavily on shell command execution to perform its primary auditing and synchronization functions.
Shell Utilities: Uses standard Unix tools (jq, grep, awk, comm, find, sed, ls, etc.) to parse metadata and validate file structures.
Local Script Execution: Triggers execution of local scripts such as scripts/slop_score.py and scripts/capabilities-sync-check.sh to conduct specialized audits.
File Operations: Includes functionality to merge, split, and delete (rm) files during the documentation consolidation phase.
[EXTERNAL_DOWNLOADS]: The skill uses the uv tool to manage environment dependencies, which may download utility packages.
Package Source: Fetches the pyyaml package from standard registries during the execution of the slop detection script.

doc-updates