Instructions for AI coding agents (Claude Code, Aider, Cursor, Continue, etc.) working in this repository.
This file is the entry point. For full details, read the files it points to.
CLAUDE.md— complete project instructions: methodology, git workflow, commit conventions, storage model, quality gates, and hard rules. Read this first.docs/README.md— documentation index with cross-references to methodology, operations, schemas.forgeplan health— run this in the terminal to see current project state (blind spots, orphans, stale artifacts).
Forgeplan — Rust-based methodology engine (CLI + MCP server + future Desktop app) for managing engineering artifacts (PRD, RFC, ADR, Epic, Spec, Evidence) with quality scoring, semantic search, and decision tracking.
- Language: Rust 1.75+ (crates workspace)
- Storage: Markdown files in
.forgeplan/as source of truth (ADR-003), LanceDB as derived index - Distribution: cargo-dist binaries, brew formula, install script
- Website: Astro + Starlight at
website/(seewebsite/README.md)
-
Follow the Forgeplan methodology itself when making non-trivial changes:
forgeplan route "task"→ determine depth (tactical / standard / deep / critical)forgeplan new <kind>→ create artifact for Standard+ depthforgeplan validate→ must PASS before codingforgeplan reason→ ADI reasoning (mandatory for Deep+)- Code → test each
pub fnimmediately forgeplan new evidence+ link + score + activate
-
Never commit to
mainordevdirectly. Always feature branch → PR. -
Never delete
.forgeplan/withoutforgeplan exportfirst. -
Never push
--forcetomain. The safety hook blocks this. -
cargo fmt+cargo checkbefore every commit. Git hooks enforce this. -
Write tests for every new
pub fnimmediately — do not move to the next function without a test. -
Markdown files in
.forgeplan/are the source of truth (per ADR-003). The LanceDB index in.forgeplan/lance/is derived — rebuild viaforgeplan scan-importif needed.
ForgePlan/
├── CLAUDE.md, AGENTS.md, README.md
├── crates/ ← Rust workspace (core + cli + mcp)
├── .forgeplan/ ← artifact workspace (markdown tracked, lance/cache/config local)
│ ├── adrs/, rfcs/, prds/, epics/, specs/
│ ├── evidence/, problems/, solutions/, notes/
│ ├── lance/ ← gitignored (derived)
│ └── config.yaml ← gitignored (local)
├── docs/ ← production documentation
│ ├── README.md ← documentation index
│ ├── methodology/ ← how to use Forgeplan
│ ├── operations/ ← agent hooks, enforcement, repo protection
│ └── schemas/ ← formal artifact schemas
├── templates/ ← markdown templates for each artifact kind
├── website/ ← official website (Astro + Starlight)
├── marketplace/ ← plugin marketplace (plugins + skills)
├── scripts/ ← build + release + helper scripts
├── Formula/ ← Homebrew formula
└── .local/ ← gitignored — local notes, research, sessions
- Documentation & commit bodies: Russian preferred (matches project conventions)
- Code identifiers & commit descriptions: English
- Communication with the user: Russian
Forgeplan is a single-author project. When generating ANY author-attributed content, use:
- Name:
Eli Rum - URL:
https://elirum.me
This applies to:
- Blog post
AuthorBlock(website/src/components/blog/AuthorBlock.astro) - Blog footer author line (
website/src/components/blog/BlogFooter.astro) - JSON-LD
author.name+Person.url(website/src/components/blog/SeoMeta.astro) - Schema.org
Organization.founder(website/src/components/SiteJsonLd.astro) - OpenGraph
article:authormeta tags package.jsonauthorfield (if updated)- Any author byline in
docs/,.forgeplan/notes/, blog.mdxfrontmatter - Conventional commit
Co-Authored-Bylines for solo work (do not add — Eli Rum is sole author)
DO NOT use placeholder names: Mike Kubal, Forgeplan Author, Anonymous, Maintainer, etc.
If a future contributor joins, this section will be updated to reflect multi-author conventions. Until then, treat the single-author invariant as load-bearing for credibility, SEO authority signal, and brand consistency.
CLAUDE.md— full project instructions (primary)docs/README.md— documentation indexdocs/methodology/FORGEPLAN-GUIDE.md— full methodology referencedocs/operations/AGENT-ENFORCEMENT.md— agent rules and guardrailswebsite/README.md— website architecture notes