carbon-monitor-adapt-harness
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads various documentation and script files from the local repository to populate sections of a generated specification, creating a potential surface for instructions embedded in those files to influence agent behavior.
- Ingestion points: Multiple files including
docs/HARNESS.md,docs/ARCHITECTURE.md,docs/GITHUB_ACCESS.md, and project scripts (scripts/rebuild-dashboard.js,scripts/init-index.js,src/api/healthChecks.js) as defined inreading-map.md. - Boundary markers: Absent; there are no explicit instructions telling the agent to ignore any natural language instructions found within the files it is reading.
- Capability inventory: The agent has the ability to read files, write the specification file (
specs/<slug>/SPEC.md), and execute shell commands (git rev-parse). - Sanitization: Absent; the skill does not specify any sanitization or validation logic for the content retrieved from the repository files before using it to generate output.
- [COMMAND_EXECUTION]: The skill requires the execution of a git command to populate the header of the generated specification file.
- Evidence:
SKILL.mdinstructs the agent to rungit rev-parse --short HEADto obtain the commit SHA. This is a routine, read-only operation typical for development workflows.
Audit Metadata