The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains a "MANDATORY" workflow section that instructs the agent to spawn three specific sub-agents in parallel (fuse-ai-pilot:explore-codebase, fuse-ai-pilot:research-expert, mcp__context7__query-docs) before any implementation. This prescriptive language attempts to override the agent's default operational logic with a specific multi-agent strategy.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow involving automated codebase exploration, which creates a vulnerability surface for indirect prompt injection.
Ingestion points: The mandatory workflow in SKILL.md requires the agent to run fuse-ai-pilot:explore-codebase, which entails reading and processing arbitrary project files.
Boundary markers: The instructions lack explicit delimiters or warnings to ignore potential malicious instructions embedded within the source code or test files being analyzed.
Capability inventory: The skill provides templates and instructions for executing powerful shell commands, including package installation (npm install, bun install) and test execution (bun test, vitest run), which involve subprocess execution.
Sanitization: No sanitization or validation logic is specified for the ingested file content before it influences the agent's implementation decisions.
[EXTERNAL_DOWNLOADS]: The CI/CD templates provided in references/templates/bun-setup.md and references/templates/vitest-setup.md reference external GitHub Actions for environment setup. Specifically, they utilize actions/checkout@v4, actions/setup-node@v4, and oven-sh/setup-bun@v2. These resources are maintained by trusted organizations and represent standard development practices.

ts-testing