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.
git clone --depth 1 https://github.com/pablomarin/claude-codex-forgeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/pablomarin/claude-codex-forge/fix-bug)<a href="https://agentmods.dev/commands/pablomarin/claude-codex-forge/fix-bug"><img src="https://agentmods.dev/badge/commands/pablomarin/claude-codex-forge/fix-bug/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/pablomarin/claude-codex-forge/fix-bug"><img src="https://agentmods.dev/badge/commands/pablomarin/claude-codex-forge/fix-bug.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.03240 |
| Opus 5.5 | $0.00000 | $0.01296 |
| Sonnet 5.5 | $0.00000 | $0.00648 |
| Haiku 4.5 | $0.00000 | $0.00324 |
Grade A, and why
fix-bug 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 10d 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- new-feature — 86% identical, 244 lines differ
How it starts
The opening of the file, as written. The whole thing — 233 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/fix-bug — Host-Neutral Systematic Bug-Fix Workflow
Diagnose and fix a reproducible defect through an open PR. The active Claude Code or Codex host is the main agent for this session; there is no permanent main engine.
Apply the shared Startup boundary before the first workflow step. It is the canonical activation, resume, isolation, and setup-failure contract.
0. Resume or Start
- Read
.forge/instructions.mdand.forge/rules/, resolve the active host, then run.forge/hooks/lib/workflow-state.sh show(PowerShell: the.ps1twin). If canonical state is absent or invalid, stop and use the setup/migration path; never reconstruct it ad hoc. - If this worktree already has the matching workflow active, use
workflow-state.sh checkpointwith the displayed phase and exact next step to recordLast active host, then resume that exact next unchecked durable step. If it is inactive, continue only through the deterministic worktree/base preflight below. Forge provides no edit lock: concurrent sessions are allowed, including simultaneous editing; coordinate overlapping writes. If any session mutates the candidate, candidate-bound evidence becomes stale and must be regenerated before certification. - Work outside the protected default branch in one isolated physical worktree, reusing an existing
prepared native worktree when present. If none exists, use the portable helper where allowed:
.forge/hooks/lib/worktree-lifecycle.sh create --kind fix --name <slug> --base <ref-or-sha>(PowerShell:worktree-lifecycle.ps1 -Action Create -Kind fix -Name <slug> -Base <ref-or-sha>). In a fresh host-native worktree, normalize and seed it before activation with.forge/hooks/lib/worktree-lifecycle.sh adopt --kind fix --name <slug> --base <ref-or-sha> --worktree "$PWD"(PowerShell:worktree-lifecycle.ps1 -Action Adopt -Kind fix -Name <slug> -Base <ref-or-sha> -Worktree $PWD.Path). This safely replaces a clean, unpublished host-generated branch name withfix/<slug>; it never renames protected, dirty, shared, or published work. Only in the Forge source checkout, when the installed path is absent, use the trackedhooks/lib/worktree-lifecycle.shor.ps1. Native creation or adoption is optional only under the shared startup boundary; never create a second worktree for the same-directory handoff. - Continue in the current session by default; switch only if the developer chooses to.
A session opened in the primary checkout may continue the linked worktree by using it as the
working directory for local edits, reviews, verification, approval recording and
normal receipt-validating promotion. Before direct commit, push or PR creation, follow
Shipping from the current session.
Shipping hooks use the host's event
cwd; selecting only the tool working directory does not change that hook context. The procedure verifies the exact task locally while preserving native hooks and authorization. The installed adapter declares the current host for reviewer routing. No per-worktree Forge receipt, copied session identity, or hook-trust bypass is required. - The helper seeds only
## State(with### Nowcleared),## Open Questions, and## Blockersfrom the primary checkout and writes the exact baseline to.forge/local/.state-seed-snapshot.md. It never seeds workflow, goal, authorization, receipts, evidence, or local memory. Thecreateandadoptactions both seed state. If an inactive native worktree already has both state and its snapshot,adoptpreserves them; if only one exists or a workflow is active, stop rather than guessing or overwriting it. - In the target worktree, run
workflow-state.sh showand resolve the intended base ref. If state is inactive, invoke.forge/hooks/lib/workflow-state.sh activate --host <claude|codex> --workflow fix-bug --task <slug> --base-ref <ref-or-sha> --phase diagnosis --next-step 'reproduce the symptom'before any discretionary investigation or tracked mutation. On Windows, use the.ps1twin. The helper resolves and freezes the base SHA. Reuse an already-recorded base for an adopted active worktree; when ancestry is ambiguous, require an explicit base before activation. Never recompute from a later-moving default branch. - Replace the active workflow checklist with:
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.
- 10d ago Changed · +14 lines 58db220f0f29
- 11d ago Changed · +3 lines aef8ff5bf29d
- 13d ago Changed · +1 lines e8fcab74ae32
- 16d ago Changed · +8 lines 644132b499af
- 17d ago Changed · +11 lines cd7de36220c4
- 20d ago Changed · +26 lines 5a99bde9640a
- 1mo ago Changed · -1,069 lines scan B → A 522c50bdc000
- 1mo ago First seen · 1,239 lines · 0 tokens per session scan B edd7ded1bf98
fix-bug is a command published in the GitHub repository pablomarin/claude-codex-forge (5 stars, last pushed 9d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,240 tokens. 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.
Other commands, from other repositories
qa-changes
This skill should be used when the user asks to "QA a pull request", "test PR changes", "verify a PR works", "functionally test changes", or when an automated workflow triggers QA validation of code changes. Provides a structured methodology for setting up the environment, exercising changed behavior, and reporting…
react-native
Command "react-native" from getsentry/toolkit, covering examples, also associate with a release and distribution(s), xcode build step (react-native xcode) and important notes.
doctor
Diagnosticar y reparar problemas del framework Don Cheli, git y entorno. Usa cuando el usuario dice "doctor", "problemas del framework", "don cheli no funciona", "repair Don Cheli", "debug setup", "setup broken", "framework broken", "reparar entorno". Detecta y repara issues de configuración, git y dependencias…
doctor
Badi configuration validation. Checks all Badi components and produces a diagnostic report.
http-service
Build, review or debug a Bun HTTP service. Loads the http-service skill, then works the task through its workflow.
auditar-cascading
Invoca cascading-failures-auditor — audita serviço para 5 triggers de cascade (sem timeout, retry sem jitter, sem circuit breaker, sem deadline, queue unbounded) e gera CASCADING-AUDIT.md.