launchdarkly-flag-drift

Pass

Audited by Gen Agent Trust Hub on Jul 28, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external API responses and local source code, creating an attack surface where malicious content could influence agent actions.
  • Ingestion points: Fetches feature flag configurations using the LaunchDarkly MCP server (get-flag tool) and scans repository source code files for flag keys and default values.
  • Boundary markers: There are no specific instructions or delimiters provided to ensure the agent ignores potential instructions embedded within flag values or code comments during analysis.
  • Capability inventory: The skill can modify local files, execute shell commands (e.g., linters, tests, codegen) discovered in the repository, and perform git operations to push branches and open pull requests.
  • Sanitization: The skill lacks explicit steps for sanitizing or validating retrieved flag configurations or code content before using them to generate code changes or pull request descriptions.
  • [COMMAND_EXECUTION]: The skill directs the agent to execute shell commands found within the target repository's configuration files.
  • Evidence: The workflow in 'Step 6' instructs the agent to discover and run formatting, linting, and testing scripts from files like package.json or Makefile. Additionally, 'Step 5' specifies running project-specific codegen commands if the defaults reside in generated files.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 28, 2026, 07:34 PM
Security Audit — agent-trust-hub — launchdarkly-flag-drift