DataArt/artisyn-implementation-harness / skills/implementation-harness
implementation-harness
FlaggedInstall the portable AI-dev harness, rebuild its operational graph standalone, harvest project skill improvements into its canonical templates, package direct canonical edits into a portable bundle, or apply canonical updates to an existing project. Use /implementation-harness launch, /implementation-harness reindex-graph, or /implementation-harness patch harvest|prepare|apply.
canonical/docs/Engineering Guide/04 AI Coding Tool Compatibility.md
View on GitHub# AI Coding Tool Compatibility
The bootstrap package installs a project-specific harness for one or more of Claude, Codex, and Copilot.
Its canonical manifest maps the same source skills and role definitions to native files.
Installed files are editable project artifacts; harvest/apply reconciles later changes.
There is no intermediate project `.harness/skills` source or continuous generator.
| Artifact | Claude | Codex | Copilot |
| ------------ | ------------------------------ | -------------------------------- | ------------------------------- |
| Instructions | CLAUDE.md | AGENTS.md | .github/copilot-instructions.md |
| Skills | .claude/skills/<name>/SKILL.md | .agents/skills/<name>/SKILL.md | .github/skills/<name>/SKILL.md |
| Agents | .claude/agents/<name>.md | .codex/agents/<name>.toml | .github/agents/<name>.agent.md |
| Settings | .claude/settings.json | .codex/config.toml | .github/copilot/settings.json |
| MCP | .mcp.json | MCP tables in .codex/config.toml | .mcp.json |
Shared documentation is written once for every selected provider. Where a doc shows a
command, it uses a leading skill-invocation prefix — `/` in Claude/Copilot, `$` in Codex — and
where it needs the root instruction file it names the one this install produced. Skill
companions (patterns, references, runtime procedures) are always referenced by their
path _inside_ the installed skill bundle, so the table above is the only place a
provider-specific root path appears.
Multi-provider installs derive shared project facts once and render provider-specific paths
and invocation syntax. Native agent models and permissions are not interchangeable.
Role instructions do not create enforced tool allowlists. Existing provider configuration
is reconciled rather than overwritten, and local credentials remain local. Claude and
Copilot share one MCP file: reconcile changes once and verify both installed consumers.
Copilot tool lists and instruction-inheritance flags depend on host capabilities;
verify the actual delegated context and tool scope rather than relying on file syntax.
## Workflow contract and runtime procedures
Skills carry selected runtime procedures for planning, tracking, interaction, delegation,
context recovery, browser inspection, and credential setup. The procedures inspect actual
session capabilities; installing a skill does not enable every native feature.
- Tasks: workflow reports own durable progress; native tracking mirrors them when available.
- Planning: explicit proposal and persisted-plan sign-offs survive either native Plan mode
or the documented planning-only fallback. Respect host write restrictions.
- Delegation: preserve role scope and independent review. If nested delegation is unavailable,
the orchestrator dispatches the reviewer. Self-review is never labelled independent review.
- Context: persist a checkpoint and re-read artifacts on resume. Compaction is optional.
- Browser: check navigation, console, network and DOM capabilities separately. Record missing
verification rather than claiming a screenshot proves an unavailable check.
Bootstrap installation state lives in the project's `.harness/state/launch/`.
Story/design reports remain the cross-provider workflow state. Package updates preserve local
launch state, which is excluded from release archives. Other coding tools require a separately
implemented and verified adapter; this package does not claim support for them.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 470k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |