ln-31-performance-optimizer
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements industry best practices for safe automation, including the requirement to isolate experiments in separate Git branches or worktrees to prevent accidental modification of the user's main workspace.
- [SAFE]: The use of a 'run-owned resource ledger' is a strong security and operational control. It mandates tracking every absolute path, process ID, and temporary artifact created during the session, ensuring a deterministic cleanup process that avoids deleting pre-existing user data.
- [INDIRECT_PROMPT_INJECTION]: As a tool designed to analyze and act upon codebases, the skill inherently possesses an attack surface for indirect prompt injection through the ingestion of external repository data.
- Ingestion points: The skill reads repository-level instructions, Git history, benchmark configurations, and profiler output (SKILL.md).
- Boundary markers: The skill does not explicitly define delimiters for external data, but it requires that experiments be verified against hard constraints (correctness, resource safety).
- Capability inventory: The skill utilizes file system access, Git command execution, and profiling/benchmark subprocess execution.
- Sanitization: No specific sanitization of input data is defined, relying instead on the agent's internal safety filters and the requirement for measurable, evidence-based outcomes.
Audit Metadata