add-agenteq-to-boost-project

Pass

Audited by Gen Agent Trust Hub on Sep 27, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill uses npx agenteq to initialize and synchronize the environment, which involves downloading and executing code from the npm registry. It also suggests running npx skills add with a remote GitHub URL for alternative refactoring tasks.
  • [COMMAND_EXECUTION]: The skill performs several shell-level operations, including running git rm -r --cached to untrack files and modifying project configuration files. It specifically targets composer.json and git hooks (e.g., Husky, pre-commit) to insert npx agenteq sync commands, ensuring the tool runs automatically during project lifecycle events.
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a reconciliation process that reads "hand-authored project guidance" from files like CLAUDE.md, AGENTS.md, and .clinerules, as well as command definitions from per-agent directories. This content is merged into a central .ai/ directory without sanitization or explicit boundary markers, potentially allowing instructions embedded in those files to influence the agent's behavior.
  • Ingestion points: Native agent guidelines files (e.g., CLAUDE.md, AGENTS.md) and native command directories (e.g., .claude/commands, .cursor/commands).
  • Boundary markers: None used for the merged content; the skill only filters specific XML tags related to Laravel Boost.
  • Capability inventory: Shell execution (npx, git), file system writes to project root and configuration files (composer.json, git hooks).
  • Sanitization: No sanitization or escaping of the hand-authored text is performed during the merge process.
  • [EXTERNAL_DOWNLOADS]: The instructions reference fetching external code and skills from https://github.com/sunchayn/agenteq, which is the vendor's repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 27, 2026, 09:47 AM