forge

A set of rules for projects hosted on a Git forge, an online service such as GitHub or GitLab that stores code and manages pull or merge requests. It defines how branches, requests, reviews, merge approval, and project records should relate.

In plain words
What is it for?
Use it when opening, updating, reviewing, or merging a pull or merge request. It helps map branches to requests, write self-contained request descriptions, perform forge reviews, obtain merge consent, and audit the contribution process.
Why use it?
It keeps the public contribution record clear and consistent with the project's internal record. It prevents private working notes or process details from leaking into public pull-request text.

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/nrdxp/predicate/forge
Any agent
npx skills add nrdxp/predicate --skill forge
Clone the repo
git clone --depth 1 https://github.com/nrdxp/predicate

Made for: Claude Code, Codex.

Per session 119 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,469 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 $0.00119 $0.03469
Opus 5 $0.00060 $0.01734
Sonnet 5 $0.00024 $0.00694
Haiku 4.5 $0.00012 $0.00347

Measured yesterday against content hash a51b7a027b2d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

forge 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 yesterday.

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.

skills/forge/SKILL.md · 274 lines

How it starts

The opening of the file, as written. The whole thing — 274 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Forge Discipline

Applicability is conditional. Some projects have no forge — a bare remote, a patch workflow, or no remote at all — and nothing here applies to them; absence of a forge is never a violation. When the project has a forge, this discipline applies in full. The test is mechanical: does the origin remote resolve to a forge the project accepts contributions through?

Philosophy: the forge is the public half of the record

The flight recorder (.ledger/) is the internal durable record; the forge is the external one — the only record most future readers will ever see. The two have different audiences and MUST NOT leak into each other: the recorder may reference forge artifacts (they are public, permanent URLs); forge prose MUST NEVER reference process-internal state. A PR body that cites .scratch/, .ledger/, a session link, or a working note is broken for every reader outside the process that wrote it — and those are all of its readers.

0. Tooling: MCP for forge operations, git for merges

Forge operations — opening/updating PRs, posting comments and reviews, reading state — prefer the forge's MCP server where the session has one connected: structured, typed calls over stringly CLI output. The CLI (gh, glab, …) is the fallback, and plain git+HTTPS the primitive beneath both (degrade-to-the-primitive applies: never depend on the MCP being present). The one hard exception runs the other way: merges are always the git CLI (§4, MERGE_IN_GIT) — no MCP or API merge endpoint, ever.

1. Branch ↔ PR mapping

Two PR tiers, one campaign:

  • Per-node PRs. Each finished node branch surfaces as a PR targeting the campaign's shared integration branch — that PR is the node's review record (findings, triage, the merge-consent trace). The merge itself still happens in git (§4); pushing the merged shared branch is what closes the PR.
  • One tracking PR per campaign. The campaign/<TOPIC> integration branch surfaces as a single PR against the default branch; independent workstreams get independent branches and PRs, never one omnibus. Its body is the campaign checklist (§2).
  • The bootstrap order is forced by the forge — follow it, never fight it. A PR cannot exist until its branch differs from its base, so:
    1. At campaign setup, create the shared integration branch from the baseline and PUSH it (covered by the campaign's recorded push authorization — preflight it with the rest). Open NO PR yet.
    2. When the FIRST node finishes, push its node branch and open the first node PR, targeting the shared branch.
    3. After that node's consented merge lands on the shared branch and is pushed, the shared branch has content: NOW open the tracking PR (draft) with its checklist body, the first item already checked.
    4. Every later node repeats the rhythm of (2)–(3): node PR → consented merge → its checklist item checked. Opening the tracking PR at campaign start — before anything exists to merge — is the recurring field failure this sequence exists to kill.
  • Stacked PRs are permitted for dependent work: base each PR on its prerequisite's branch. Know the forge's mechanics: when a base branch merges and is deleted, the dependent PR retargets (or auto-closes as "closed", not "merged") — when the forge misclassifies, annotate the PR with the merge commit so the record reads correctly.
  • Merged branches are deleted; the PR is the durable pointer.

Read the full file on GitHub · 274 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. yesterday First seen · 274 lines · 119 tokens per session scan A a51b7a027b2d

Subscribe to this mod's changes

forge is a skill published in the GitHub repository nrdxp/predicate (10 stars, last pushed 8d ago), licensed MIT. It adds 119 tokens to every session and 3,469 once invoked, about $0.0006 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

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 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

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens