forge-core

Core rules for Forge, a local-first command-line tool for running programmable software-engineering agents. They define required verification steps, protected code areas, permission checks, sandboxing, and credential storage.

In plain words
What is it for?
Developing or reviewing Forge changes, especially its agent loop, execution, validation, model routing, persistence, permissions, filesystem sandbox, shell-risk checks, keychain use, formatting, linting, builds, and tests.
Why use it?
They reduce the risk of unsafe tool use, unauthorized file access, insecure credentials, invalid state changes, and regressions in the Forge codebase.

Cursor rule for Cursor

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 rules/hoangsonww/forge-agentic-coding-cli/forge-core
Clone the repo
git clone --depth 1 https://github.com/hoangsonww/Forge-Agentic-Coding-CLI

Made for: Cursor.

Per session 601 This file is loaded in full into every session.
When invoked 601 The same file — it is already loaded in full.
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 $0.00601 $0.00601
Opus 5 $0.00300 $0.00300
Sonnet 5 $0.00120 $0.00120
Haiku 4.5 $0.00060 $0.00060

Measured yesterday against content hash c0eea1133b0c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

forge-core 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.

.cursor/rules/forge-core.mdc · 70 lines

How it starts

The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Forge — core conventions

Forge is a local-first, multi-agent, programmable software-engineering CLI runtime (TypeScript, Node 20+). Ships as @hoangsonw/forge on npm and as a multi-arch Docker image.

Canonical verify chain

Before claiming any change done:

npm run format && npm run lint && npm run build && npm test

249 tests across 43 files must remain 100% passing.

Hot paths — surgical edits only

  • src/core/loop.ts — agentic pipeline
  • src/agents/executor.ts — iterative tool-use loop
  • src/core/mode-policy.ts — per-mode caps
  • src/core/validation.ts — post-step gate
  • src/models/router.ts + src/models/adapter.ts — routing
  • src/persistence/tasks.ts — state machine + LEGAL_TRANSITIONS

Non-negotiables

  • Every tool call goes through requestPermission (src/permissions/manager.ts).
  • Paths resolved to realpath and confined to project root (src/sandbox/fs.ts).
  • Shell commands classified by classifyCommandRisk (src/sandbox/shell.ts); critical is hard-blocked.
  • Credentials in OS keychain (src/keychain/), never in config files.
  • State transitions live in LEGAL_TRANSITIONS — never monkey-patch.

Style

  • TypeScript strict. No any in production without a comment explaining why (exception: src/tools/registry.ts).
  • Comments explain why, not what. Never add a comment that describes the change you just made.
  • Prefer readonly and immutable data flow. Prefer function modules over classes (exception: provider classes in src/models/).
  • Throw ForgeRuntimeError (src/types/errors.ts) with { class, message, retryable, recoveryHint? }. Use Result<T, E> shapes for expected failures.

Performance posture

  • REPL cold-start < 250 ms.
  • forge doctor < 1 s with no providers running.
  • UI app.js < 120 KB uncompressed.
  • No synchronous disk reads on REPL redraw or UI poll paths.
  • Provider availability probes ~1.5 s, not long timeouts.

Before adding a dependency, ask whether a 20-line hand-written utility would do. It usually would.

Read the full file on GitHub · 70 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 · 70 lines · 601 tokens per session scan A c0eea1133b0c

Subscribe to this mod's changes

forge-core is a cursor rule published in the GitHub repository hoangsonww/Forge-Agentic-Coding-CLI (22 stars, last pushed 15d ago), licensed MIT. It adds 601 tokens to every session, about $0.0030 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 2026-08-30.