The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local initialization scripts (scripts/bash/setup-change-publish.sh and scripts/powershell/setup-change-publish.ps1). These scripts are used to resolve directory paths and count local files using standard system utilities like git, grep, and basename. The scripts are self-contained and do not perform network operations.
[INDIRECT_PROMPT_INJECTION]: The skill ingests content from local markdown files (ChDR-*.md) to generate records in the docs/ directory. This creates a surface where malformed input could influence documentation output, though the impact is localized to project documentation.
Ingestion points: Draft files located in .adlc/drafts/chdr/.
Boundary markers: The skill does not use specific delimiters when wrapping the decision content in new files.
Capability inventory: File system read/write access; no network capabilities are requested or used.
Sanitization: Includes a validation step to ensure decision claims are backed by commit SHAs or URLs, providing a basic integrity check.
[SAFE]: No high-severity security threats such as credential exfiltration, remote code execution from untrusted URLs, or obfuscated payloads were identified in the skill instructions or scripts.

change-publish