3_fix

A repair workflow for addressing problems documented in a code review report.

In plain words
What is it for?
It reads the review and related specifications, fixes issues by severity, checks project commands, and stops when a recurring problem needs user guidance.
Why use it?
It turns review findings into a controlled fix process while keeping changes within the current project requirements and phase.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/dixus/claudeframework/3_fix
Any agent
npx skills add dixus/claudeframework --skill 3_fix
Clone the repo
git clone --depth 1 https://github.com/dixus/claudeframework

Made for: Claude Code, Codex.

Per session 27 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,206 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00027 $0.01206
Opus 5 $0.00014 $0.00603
Sonnet 5 $0.00005 $0.00241
Haiku 4.5 $0.00003 $0.00121

Measured 2d ago against content hash b1b6e7c83b85, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

3_fix scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/3_fix/SKILL.md · 68 lines

How it starts

The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Fix issues found in a review report.

Steps:

  1. Read CLAUDE.md — note the project's test command, lint command, build command, and typecheck command if listed

  2. If $ARGUMENTS is provided, read .claude/reviews/<name>-review.md. Otherwise read the most recently modified file in .claude/reviews/

  3. If a corresponding spec exists in .claude/specs/, read it too — fixes must stay within the spec's stated requirements a. Phase awareness: if .claude/specs/<name>-phases.md exists, read it. Fixes must stay within the current phase's scope. If a review issue references a requirement that belongs to a later (pending) phase, skip it with "out of scope for current phase" — do not implement it.

  4. Parse the Issues section — work through them in order of severity: critical first, then major, then minor

  5. Circuit breaker: for each issue, check whether it appeared in a previous review of the same spec. If the same issue has recurred across two or more review cycles, flag it as a recurring failure, stop fixing, and escalate to the user with a description of the root cause — do not attempt a third fix for the same problem without user guidance.

  6. Scope gate: only touch files that are explicitly mentioned in the Issues section of the review report. For each issue: read the affected file, understand the problem, then apply a minimal fix — do not refactor, clean up, or improve any code outside those files. Unrelated issues belong in the next review cycle, not this fix cycle.

  7. After fixing all issues, run the project's verify commands in this order (read them from CLAUDE.md, skip any not listed): a. Typecheck b. Lint c. Tests d. Build (if listed)

  8. If any verify step fails, fix those failures too. If a verify step still fails after two fix attempts, stop and escalate to the user — do not loop.

  9. Validation criteria check: if the spec has a "Validation criteria" section, confirm each criterion is still met after the fixes — list each criterion and its status. Any failing criterion is a new critical issue that must be fixed before continuing. (In phased mode, only check criteria assigned to the current phase.)

  10. Self-check against review items: before declaring done, re-read the review file and verify each issue one by one: a. For each issue marked in the review, re-read the affected file at the specific line b. Confirm the fix actually addresses the issue described — not just a nearby change c. Check that the fix didn't introduce a new problem in the same area (e.g. breaking an import, changing a signature without updating callers) d. If any issue is NOT actually resolved, fix it now before proceeding This step prevents the common pattern where /3_fix reports "all fixed" but the next /2_review finds the same issues still present.

  11. Spec-anchored check: review whether any fixes diverged from the spec's stated requirements. If they did, do not update the spec — the spec is ground truth. Instead, document the divergence in the report (step 13) and flag it for user decision. Only the user may approve spec changes.

  12. Self-improvement gate: for each fixed issue, ask whether a rule would prevent recurrence. If yes, write it to .claude/context/lessons.md immediately using this format:

    ## YYYY-MM-DD — <topic>
    **scope:** framework | project
    
    **What went wrong**: ...
    **Rule**: ...
    

    Scope guidelines:

    • scope: framework — the lesson applies to any project using this framework (e.g., "never skip typecheck", "circular import prevention"). These can graduate to CLAUDE.md rules via /ship Step 4b.
    • scope: project — the lesson is specific to the current codebase (e.g., "HelpSection goes below panel header", "Likert zero-value bug"). These stay in lessons.md and are stripped by /deploy.

    When in doubt, use scope: project — it's safer to under-promote than to pollute the framework with project-specific rules. Also flag it in the report so the team can consider promoting it to CLAUDE.md.

Read the full file on GitHub · 68 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 68 lines · 27 tokens per session scan A b1b6e7c83b85

Subscribe to this mod's changes

3_fix is a skill published in the GitHub repository dixus/claudeframework (10 stars, last pushed 4mo ago), licensed MIT. It adds 27 tokens to every session and 1,206 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

brainstorming

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

obra/superpowers · 37 tokens

chat-pet-sprite-creation

Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.

microsoft/vscode · 53 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

agent-host-chat-contributions

Build and review cross-cutting agent-host chat behavior through lifecycle contributions. Use when adding turn lifecycle side effects, prompt or context injection, restored-history transformation, protocol-action observation, or when reviewing changes that add code to AgentSideEffects or AgentService.

microsoft/vscode · 56 tokens

auto-perf-optimize

Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.

microsoft/vscode · 62 tokens