docs-impact-architect
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection Surface]: The skill ingests untrusted data from external sources, specifically GitHub Pull Request diffs and descriptions. This presents a potential surface where adversarial instructions embedded within a PR could attempt to influence the agent's output. However, the skill's capabilities are focused on generating structured documentation plans (TOC deltas) and JSON reports, which limits the potential impact of such attempts. For review purposes, the ingestion points are the PR diff and description, while the skill's capabilities include JSON emission and local file reads.
- [Command Execution]: The instructions direct the agent to use the GitHub CLI tool (
gh pr diff) to retrieve pull request information. This is a standard and expected use of development tools for a skill intended to analyze code changes and their impact on documentation. - [Local File Access]: The skill reads a local documentation index file (
.apm/docs-index.yml). This access is necessary for the agent to understand the existing documentation structure and appears confined to non-sensitive project-specific metadata.
Audit Metadata