The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes local markdown files (CDRs) to generate new documentation and metadata. While this presents an attack surface where malicious content in a CDR could influence generated output, the skill incorporates a validation gate and intended review workflow.
Ingestion points: Reads CDR-*.md files from the local .adlc/drafts/cdr directory during the generation phases (Phases 5, 6, and 7).
Boundary markers: The skill uses structural markdown headers (e.g., ### Status: **Accepted**, ### Context) to delimit content, though it does not explicitly describe sanitization of the values extracted from these sections before they are placed in YAML frontmatter.
Capability inventory: The skill can write files to the filesystem and execute Git/GitHub CLI commands to commit and publish changes.
Sanitization: The skill includes a 'Signal Gate Validation' (Phase 2) that filters out records lacking concrete evidence or project-wide applicability, acting as a functional quality and safety filter.
[COMMAND_EXECUTION]: The skill utilizes shell scripts and CLI tools to manage the publication workflow.
Evidence: It executes git for branch creation, committing changes, and pushing to remotes. It also uses the GitHub CLI (gh) to automate the creation of draft pull requests. These actions are standard for development automation and are clearly documented in the skill's instructions.

levelup-publish