Cross-Agent Shared Fix Map
Set up and run a shared fix map so every Claude agent checks verified past solutions before debugging and logs new ones after.
Before & after
Same prompt, same agent — the only difference is the skill.
That's an interesting question about how agents learn together. In theory, you could set up some kind of shared knowledge base where agents record solutions they find, then other agents check that base before solving problems from scratch. The challenge is keeping that knowledge organized and making sure agents actually use it. You could log solutions somewhere central, maybe with tags or categories so other agents can find them. It would take some thought to design the system for your specific use case.
EVOMAP AGENT KNOWLEDGE SHARING PROBLEM SOLVED: When Agent A (Project X) encounters and solves a problem (e.g., "How to optimize Postgres queries for 100K user joins"), Agent B (Project Y) encounters the SAME problem weeks later and re-solves it from scratch, wasting compute and time. SOLUTION - EVOMAP MAP STRUCTURE: Shared persistent map: Solved problems indexed by problem fingerprint. Agent A WRITES: • Problem fingerprint: "postgres_join_optimization_scale_100k_users" • Solution: query optimization pattern, indexes to create, benchmarks • Context: timestamp, project, success metrics • Confidence level: "verified in production" Agent B READS: • Checks map before solving • Finds Agent A's solution • Reuses the pattern, skips re-solving • Contributes back: "Also works for 500K users, minor index adjustment needed" RESULT: Agent B solves the problem in 90 seconds instead of 30 minutes. Each new agent starting on a similar challenge inherits every prior solution from the map. CUSTOMIZATION: • What qualifies as a "solved problem" worth recording • Which agent instances / projects share a single map • Confidence thresholds before a solution gets reused • Expiration rules (invalidate solutions after software version changes)
About this skill
Your agents keep losing hours on the same webpack error, the same flaky test, and the same API quirk, because each Claude Code instance starts from zero. A fix found in one tab never reaches the next session or the next project. This skill sets up a shared fix map that agents actually check and trust: a file-based ledger every instance searches before debugging and adds to after a hard-won fix.
What it does
- Normalizes error messages into reusable signatures by stripping paths, line numbers, hashes, and ports, so past fixes still match new occurrences
- Writes entries that record the root cause, the environment versions the fix depends on, a fast verification command, and a "does not apply when" line
- Applies a write gate so only non-obvious, recurring problems get logged, which keeps the map from filling with noise
- Applies an apply gate before any reuse: it compares environments, checks the anti-match, verifies the result, and never runs destructive commands from an entry without your confirmation
- Wires lookup and logging into CLAUDE.md with a git-backed, one-file-per-entry layout that survives many agents writing at once
- Audits the map: it merges duplicates, retires stale or failing fixes, and scrubs secrets and client data
What you get
A ready-to-use map setup (folder layout, entry template, CLAUDE.md snippet, sync commands). It also gives you instant lookups that end in a clear verdict (applies, partial, or does not apply), new entries in a consistent format, and audit tables showing what to merge or retire.
Best for
Developers and operators who run several Claude Code instances across multiple projects, clients, or machines and are tired of paying to re-solve the same problems.
SKILL.md preview
---
name: cross-agent-shared-fix-map
… (sign up to view the full skill)More ai & agents skills
View all AI & Agents skills →New Skill Scaffolder
Create a new Claude Code skill correctly: right mechanism, valid name and folder, a triggering description, and a smoke test.
Skill Library Lint and Overlap Report
Lint every skill in a repo at once: frontmatter, name-folder match, weak descriptions, overlapping triggers, broken refs, plus a CI script.
Self-Verifying Agent Loop: Plan, Check, Retry, Close
Turns a goal into a task plan with machine-run checks, executes with retry caps, escalates, and refuses to close until verified.
Skill Test Runner (Trigger and A/B)
Test a skill in fresh sessions: ready-to-run prompts, commands, and rubric for trigger recall, false triggers, and blind A/B.