harneo
CLAUDE.md4 starsChanged 13 days agoMITView on GitHub
Harneo is a harness manager for Claude Code: a Tauri desktop app that keeps a library of AI Artifacts, composes profiles from it, and scaffolds or refits projects from a profile. Read CONTEXT.md before designing or building anything. It is the product source of truth: artifacts, user journey, feature decisions and constraints.
- Reads credentials
- Commits and pushes
In this repository2 other files it gives its agents.
- security-guardrailsSkill
- ticket-flowSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 25
ADRs only through Addy's framework: write one with the documentation-and-adrs skill when a phase calls for it (e.g.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Harneo
Harneo is a harness manager for Claude Code: a Tauri desktop app that keeps a library of AI Artifacts, composes profiles from it, and scaffolds or refits projects from a profile. Read CONTEXT.md before designing or building anything. It is the product source of truth: artifacts, user journey, feature decisions and constraints.
Status
- The
app-shellmodule exists: the Tauri app frame (src-tauri/,src/),harneo-coreand theharneoCLI (crates/), the quality gates (scripts/check-*.sh,CONSTRAINTS.md) and the setup script (scripts/setup.sh). Library, profile and project features are not built yet. The repo also holds the product context, specs (spec/), the design system (DESIGN.md,PRODUCT.md) and the validated clickable prototype (design/prototype/, the visual reference for every UI ticket). - Stack decided: Tauri 2 + React + Vite + TypeScript, with a shared Rust core (
harneo-core) for the app and the CLI. - Modules and build order:
spec/capabilities.md. Each module getsspec/<module-id>.mdbefore any tickets (Addy's spec → plan → tasks → build).spec/app-shell.mdalso holds the project-wide conventions: commands, structure, code style, testing, boundaries. - Cross-platform code, built and tested on Linux only. Distribution is source for every OS plus Linux packages (AppImage, .deb) attached to each release by the
linux-packagesworkflow (CONTEXT.md§4–5, ADR-003).
Commands
Setup and per-OS prerequisites: BUILDING.md (start with ./scripts/setup.sh, then npm ci).
npm run tauri dev: run the app with hot reload.npm run tauri build -- --no-bundle: release binary, no installer.cargo run -p harneo-cli -- --version: run the CLI (pathsprints the data folder).npm run check:fast: after each edit.npm run check:task: before a ticket is reported done.npm run check:full: before any merge tomain.npm test,npm run lint,npm run typecheck,npm run fix: the building blocks. Shells may needeval "$(mise env -s bash)"to see the gate tools.
Quality bar
- Read
CONSTRAINTS.mdbefore writing code. Do not weaken it to make a change pass. Every rule blocks:npm run check:taskmust pass before a ticket is reported done, andnpm run check:fullbefore any merge tomain.
Specs and decisions
- Specs live in
spec/, never at the repo root. When/spec(Addy'sspec-driven-development) writes a new spec, save it asspec/<module-id>.md, notSPEC.md.spec/capabilities.mdis the index of modules and specs. WORKFLOW.mdis the user's log of the framework: one line per step run (Addy's commands and Impeccable's), with what it produced. Add a line every time a step runs.- ADRs only through Addy's framework: write one with the
documentation-and-adrsskill when a phase calls for it (e.g. a hard-to-reverse choice while specifying a module), in that skill's default locationdocs/decisions/. Never write ADRs by hand outside that flow.
Tickets and workflow
- Follow the
ticket-flowskill (.claude/skills/ticket-flow/) for all work. No change without a GitHub issue inibrahimkobeissy/harneo. - Board: GitHub Project 5 ("Harneo", https://github.com/users/ibrahimkobeissy/projects/5), with Status Backlog → Ready → In progress → In review → Done. Config:
.claude/ticket-flow.conf. Move statuses with.claude/skills/ticket-flow/scripts/board.sh. - Trunk-based: everything lands on
main. No feature branches, no PRs. Subagents work in detached worktrees under.claude/worktrees/, and the orchestrator session validates, fixes, merges and pushes. - Tickets use
.claude/skills/ticket-flow/references/ticket-template.md, with dependencies written asBlocked by #N, one per line. Never delete a ticket; close it as "not planned" or as a duplicate. tasks/plan.mdkeeps only the ordered list of issue links, never a duplicate checklist. Do not writetasks/todo.mdor use Addy's/build auto; the ticket-flow cycle replaces them.- Releases: a release is a
release_x.y.ztag onmainwhose version matchespackage.json,Cargo.tomlandsrc-tauri/tauri.conf.json. Pushing the tag runs thecheckworkflow (.github/workflows/check.yml). Wait for that run: when it passes, publish a GitHub Release for the tag at once (gh release create release_x.y.z --verify-tag --latest --title "Harneo x.y.z" --notes-file <notes>), with notes shaped like release_0.1.0: what changed since the previous release, install (Linux packages, and source with a link to the README's per-OS steps), platform status, privacy changes, thanks. Publishing starts thelinux-packagesworkflow (.github/workflows/linux-packages.yml): wait for it and check the release lists the AppImage, the .deb andSHA256SUMS; if it fails, fix it under a ticket and re-run it by hand for the tag (gh workflow run linux-packages.yml -f tag=release_x.y.z). When the run fails, publish nothing: fix it under a ticket, check it with a manual run (gh workflow run check.yml --ref main), move the tag only with the user's approval, and release once the run on the tag passes. - Cloud sessions have no
ghand no board access: use the GitHub MCP tools for issues and record statuses asstatus:*labels, which the next local session'sboard.sh syncapplies to the board.
Security Guardrails
Agents operate inside the project directory. Treat anything outside it as
off-limits unless the user explicitly asks. The paths below are hard
prohibitions: never Read, Edit, Write, cat/less/grep/copy, or
otherwise access them, and never echo their contents into a file, a note, a
web request, or the chat (no exfiltration).
- Credential & key stores:
~/.ssh,~/.aws,~/.gnupg,~/.config/gcloud,~/.azure,~/.kube,~/.oci,~/.docker/config.json,~/.netrc,~/.git-credentials,~/.config/gh,~/.npmrc,~/.pypirc. - Agent tokens:
~/.config/anthropic,~/.claude/.credentials.json,~/.codex,~/.gemini. - Secret files anywhere:
.env/.env.*,*.pem,*.key,*.p12,*.pfx,*.jks,*.keystore,id_rsa*,id_ed25519*,id_ecdsa*,*service-account*.json,*credentials*.json,secrets.*,*.secret,*.tfstate*,.pgpass,.my.cnf,.vault-token. - Shell/DB history:
~/.bash_history,~/.zsh_history,~/.python_history,~/.psql_history.
Enforcement (Claude Code): permissions.deny in .claude/settings.json
covers the native Read/Edit/Write tools, and the PreToolUse hook
(guard-sensitive-paths.py) covers shell commands — including the cat …
case that path-based denies cannot stop. If a task genuinely requires touching
one of these paths, stop and ask the user to do it manually; do not work
around the guardrail.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

