Score agent loops against a verifiable contract and gap checklist.
npx skills add https://github.com/SollanSystems/loop-engineer --skill loop-inspector
Please help me install this Agent Skill. Skill: loop-inspector Source: https://github.com/SollanSystems/loop-engineer/tree/main/skills/loop-inspector Command: npx skills add https://github.com/SollanSystems/loop-engineer --skill loop-inspector
This Skill helps you evaluate an existing agent loop or harness and determine whether it is actually robust, verifiable, and safe against false completion claims.
Ask the skill to inspect the target loop directory and return a scored gap report with the missing defenses clearly listed.
High-intent search queries and answers about installing and using this skill.
You can audit an agent loop for false completion by scoring it against a verifiable contract and gap checklist. This read-only inspection evaluates terminal-state coverage, approval gates, and false-completion defenses to ensure the harness only reports success when independently verified.
Gap analysis for agent loop harnesses is a read-only inspection that scores an existing loop directory against a prime-directive checklist. It verifies independent verification criteria, terminal-state coverage, and false-completion defenses to determine if a long-running workflow is robust and safe to adopt.
You check terminal-state coverage in a third-party agent harness by performing a read-only inspection of named contract files. The audit verifies approval gates and false-completion defenses, producing a scored gap report that clearly lists any missing terminal states.
Agent loop verification does not require modifying existing harness code. The inspection is strictly read-only, reviewing named contract files, approval gates, and terminal-state coverage against a gap checklist without altering the inherited loop directory.
You need to run a verification audit on an inherited agent workflow before adoption to ensure it is robust and safe. Run the gap analysis when inheriting a harness, comparing loop designs, or doing a final review of long-running agent workflows to catch false-completion vulnerabilities.
XinAlohaCreate machine-verifiable JSON contracts defining agent loop objectives and constraints.
sidunrealdeAudit Nyquist validation gaps and generate missing tests for a development phase.
SollanSystemsRoutes agent-loop objectives into design, contract, run, repair, eval, or flywheel workflows.
postpo-cmykAudit project codebases for security vulnerabilities in read-only mode.
