Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/pablomarin/claude-codex-forgenpx agentmods add commands/pablomarin/claude-codex-forge/opinionWrote 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/opinion)<a href="https://agentmods.dev/commands/pablomarin/claude-codex-forge/opinion"><img src="https://agentmods.dev/badge/commands/pablomarin/claude-codex-forge/opinion/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/opinion"><img src="https://agentmods.dev/badge/commands/pablomarin/claude-codex-forge/opinion.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 | $0.02436 |
| Opus 5.5 | $0.00000 | $0.00974 |
| Sonnet 5.5 | $0.00000 | $0.00487 |
| Haiku 4.5 | $0.00000 | $0.00244 |
Grade A, and why
opinion 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 13d 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.
How it starts
The opening of the file, as written. The whole thing — 168 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Forge opinion workflow — Fresh Review and Investigation
Use the fixed-target compatibility launcher registered by the active Claude Code or Codex adapter.
It declares the current host to the dispatcher; main_host is routing metadata, not authenticated
session identity or reviewer evidence. Never invent a main engine flag. User-facing reviewer
choices map only to --engine auto|claude|codex; auto means the other engine. A healthy explicit
same-engine request is a fresh independent reviewer, not a fallback.
Inputs
Claude Code invokes this workflow as /opinion <request>; Codex invokes it as $opinion <request>.
Adding investigate after either host-native entry point selects the investigation profile.
Otherwise classify the request as one of: general second opinion/analysis/brainstorming/question
(general), plan, PRD, review comments, code review, or independent investigation reproduction.
General is hermetic and read-only. Resolve the workflow's persisted immutable base SHA/ref through
the bounded workflow-state.sh show reader; never recompute it from a moving default branch.
Put the exact request in a regular prompt file under .forge/local/reviews/.
For the ordinary review profile, FORGE_REVIEW_TRANSPORT_AUTHORIZED communicates the developer's
standing human approval in the canonical Human-Approved Reviews section (.forge/instructions.md;
source: FORGE.template.md). Send the complete bounded immutable candidate
snapshot, prompt, and evidence to the configured Claude Code or Codex reviewer service without
a separate Forge consent question, including follow-up reviews, fallback, and resumed sessions.
This approves review execution, not acceptance of findings or shipping. Private or unchanged
tracked content does not require another approval. This transport is not an external mutation and
does not grant arbitrary network tools, additional secrets or credentials, gitignored developer
state beyond the candidate, outside-worktree access, other projects, arbitrary destinations, or
external mutations. Respect host security controls and cite the actual human standing instruction
when authorization provenance is required. Include the existing human instruction in the launch
justification, together with the configured destination and bounded input. Do not ask the human to
reapprove private snapshot transport already covered by that instruction. If a host rejection omitted
existing authorization, correct the request with that evidence through the normal host mechanism;
if still denied, report the host limitation without changing reviewer scope or bypassing controls.
The same standing human approval covers full-agent investigation selected from task needs;
its different capabilities are described below. Neither mode adds a Forge launch-consent question.
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.
- 13d ago Changed · +59 lines 8f3e8d677bb5
- 16d ago Changed · -1 lines 15f5ec25e2ab
- 20d ago Changed · +6 lines 33562a500b53
- 1mo ago Changed · +17 lines 3b032dbef1c4
- 1mo ago Changed · +8 lines d541a45c4cfc
- 1mo ago First seen · 79 lines · 0 tokens per session scan A 404b1b5600e3
opinion 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 2,436 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 2026-09-04.
Other commands, from other repositories
refactor
Systematic code refactoring with smart scope detection.
review-branch
Review the current branch's diff against base by dispatching atomic-reviewer. No orchestration loop, no spec required — pre-flight before /commit pr or /commit merge.
init
Install the formatters this repository needs, with every command visible before it runs.
merge-conflict-analysis
You are analyzing merge conflicts for PR #${{ pr-number }}.
boot
Start a new PromptKit prompt assembly session (alias for /bootstrap).
repo-audit
Audit a codebase (local or remote GitHub/GitLab) against architecture principles and requirements, surfacing drift, risk, and missing decisions.