instrument-feature-flag
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by reading local application source code, manifests (package.json, requirements.txt), and lockfiles to detect frameworks and SDK versions. This provides a surface where malicious instructions embedded in codebase comments or metadata could theoretically influence the agent, although the skill's instructions prioritize structural analysis and explicit user approval for any edits.
- [REMOTE_CODE_EXECUTION]: The skill identifies missing feature flag SDKs and proposes installation via standard package managers (npm, pip, etc.). These operations target official registries and require user approval as part of the implementation plan.
- [DATA_EXFILTRATION]: The skill performs network operations to retrieve feature flag definitions and environment configurations from the Harness platform via the Harness MCP server or CLI. These operations are restricted to the intended service provider and the skill explicitly forbids the hardcoding, requesting, or printing of credential values.
Audit Metadata