resolving-merge-conflicts
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from git conflict hunks and pull request metadata to determine how to resolve code conflicts. It further instructs the agent to execute project-defined "automated checks" (tests, linters) found in the environment. This ingestion of untrusted data combined with the execution of project scripts creates a surface for indirect prompt injection.\n
- Ingestion points: Git conflict hunks, commit history, and pull request descriptions.\n
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the instructions.\n
- Capability inventory: Shell execution for git commands and the ability to run local project scripts (e.g., via npm, make, or pytest).\n
- Sanitization: No sanitization or validation of the conflict content is specified.\n- [COMMAND_EXECUTION]: Step 4 of the instructions explicitly directs the agent to "Discover the project's automated checks and run them," leading to the execution of arbitrary commands defined in the project's configuration files.\n- [EXTERNAL_DOWNLOADS]: The
UPSTREAM.mdfile contains instructions to install the skill collection usingnpxfrom the vendor's GitHub repository (LightDevCoder/skills-3rdParty).
Audit Metadata