changelog-scan
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches release notes and API documentation from official domains (code.claude.com).
- [COMMAND_EXECUTION]: Invokes a local execution of the vendor-owned package (@fusengine/harness) using the bun runtime to parse changelog data.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external documentation content, which presents a surface for embedded instructions.
- Ingestion points: External URLs listed in
SKILL.mdandreferences/sources.md(e.g.,code.claude.com/docs/en/changelog.md). - Boundary markers: The skill uses a structured report template located in
references/templates/changelog-report.mdto format findings. - Capability inventory: Execution of a local CLI tool within the agent environment as defined in
SKILL.mdand file system writes to~/.claude/logs/for state management. - Sanitization: The skill logic is designed to parse specific version headers and structural elements (e.g.,
<Update label="X.Y.Z">) rather than executing the raw content.
Audit Metadata