mcs-fix Skill
Use this skill to fix an agent after eval failures. Classifies root causes (instruction gaps, boundary violations, routing failures, knowledge gaps, scoring issues, decision mismatches), applies targeted fixes via PE and TE, then re-evaluates. Use after /mcs-eval shows failures, not during initial build (which has its Published by microsoft in MCS-Agent-Builder.
What is mcs-fix Skill?
Use this skill to fix an agent after eval failures. Classifies root causes (instruction gaps, boundary violations, routing failures, knowledge gaps, scoring issues, decision mismatches), applies targeted fixes via PE and TE, then re-evaluates. Use after /mcs-eval shows failures, not during initial build (which has its Published by microsoft in MCS-Agent-Builder. This profile combines repository metadata with install, compatibility, and usage signals so developers can quickly decide whether it fits their agent workflow before opening the source repository.
Automated repository signals based on public metadata such as recency, license, installation evidence, and adoption. These are not a security audit or endorsement.
Key capabilities
- Includes SKILL.md support
- Reusable instructions support
- Testing
- Testing use cases
Technical details
- Install or run with Copy skill directory
When to use mcs-fix Skill
- Use it for testing.
Built with
Editorial notes
Source
- Creator: microsoft
- Repository: microsoft/MCS-Agent-Builder
- Skill file: .claude/skills/mcs-fix/SKILL.md
What it does
Use this skill to fix an agent after eval failures. Classifies root causes (instruction gaps, boundary violations, routing failures, knowledge gaps, scoring issues, decision mismatches), applies targeted fixes via PE and TE, then re-evaluates. Use after /mcs-eval shows failures, not during initial build (which has its
Skill instructions
MCS Fix — Post-Eval Fix & Re-Evaluate Analyze eval set failures from /mcs-eval, classify root causes, generate and apply targeted fixes, then re-evaluate to measure improvement. Input /mcs-fix {projectId} {agentId} Reads: agentspec.json — evalSets (tests with lastResult), instructions, integrations, capabilities, conversations.topics Writes: agentspec.json (instructions, conversations.topics, evalSets, notes.fixHistory), agent in MCS (via hybrid stack) Prerequisites: Auth Verification Re-verify auth from /mcs-build. Quick silent check — az account show must match buildStatus.azTenantId, Dataverse must be reachable. If missing → "Run /mcs-build first." Step 1: Read & Validate Eval Results 1. Read agentspec.json.evalSets[] — scan for lastResult 2. No results → exit: "Run /mcs-eval first." 3. All sets passing → exit: "All eval sets passing. Nothing to fix." 4. Output per-set pass rates and failing test count Step 2: Classify Failures (Lead + QA + eval-guide plugin) Before classification —
Explore related resources
Frequently asked questions
What is mcs-fix?
mcs-fix is a open-source AI agent skill with Copy skill directory. Use this skill to fix an agent after eval failures. Classifies root causes (instruction gaps, boundary violations, routing failures, knowledge gaps, scoring issues, decision mismatches), applies targeted fixes.
Who is mcs-fix best for?
mcs-fix is best for reusing agent instructions, scripts, and references, testing workflows.
How do I install mcs-fix?
Install or run mcs-fix using Copy skill directory. Check mcs-fix for the latest setup command.
Is mcs-fix actively maintained?
mcs-fix may need a closer maintenance check before production use.
Auto-fetched from GitHub.
