DataArt/artisyn-implementation-harness / .claude/skills/evolve
evolve
VerifiedImprove harness skills, agents, workflows, and conventions from observed friction or a concrete improvement request. Supports reflect, triage, validate, and adopt; use for harness improvements rather than ordinary product changes.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 1.3 KB |
| interfaces/improvement.md | markdown | 1.2 KB |
| interfaces/reflection.md | markdown | 353 B |
| interfaces/validation.md | markdown | 653 B |
| patterns/task-tracking.md | markdown | 777 B |
| references/00-router.md | markdown | 1007 B |
| references/01-reflect.md | markdown | 805 B |
| references/02-triage.md | markdown | 2.6 KB |
| references/03-draft.md | markdown | 1.2 KB |
| references/04-validate.md | markdown | 1.5 KB |
| references/05-adopt.md | markdown | 1.5 KB |
| runtime/browser.md | markdown | 2.5 KB |
| runtime/context.md | markdown | 517 B |
| runtime/credentials.md | markdown | 512 B |
| runtime/delegation.md | markdown | 910 B |
| runtime/interaction.md | markdown | 359 B |
| runtime/planning.md | markdown | 745 B |
| runtime/review.md | markdown | 1.0 KB |
| runtime/task-tracking.md | markdown | 897 B |
runtime/browser.md
View on GitHub# Browser verification
Use Claude-in-Chrome when its tools are exposed and authorized.
Check capabilities separately: navigation, interaction, network inspection, console
inspection, and DOM/accessibility inspection. Follow the integration's tool rules.
Do not replace missing network or DOM evidence with a screenshot claim.
Preserve the skill's origin scope and explicit opt-in. Report unsupported checks
as not verified; optional browser checks never block the rest of the review.
## When the tools are not exposed
Do not silently downgrade the check. Tell the developer what is missing, walk them
through the setup below, and continue only once they confirm. If they decline or the
integration cannot be enabled, record the browser checks as not verified and move on.
Requirements: a Chromium-based browser; a direct Anthropic plan (Pro, Max, Team or
Enterprise — not Bedrock, Vertex AI or Foundry); a `/login` session, because API keys
and long-lived tokens disable the integration; a current Claude Code version. Chrome
integration is not supported under WSL.
1. **Install** — add the Claude extension from the Chrome Web Store and confirm it is
enabled in `chrome://extensions`, or start a session with `claude --chrome` and
choose **Install extension** at the prompt.
2. **Connect** — run `claude --chrome`. If the connection drops mid-session, run
`/chrome` → **Reconnect extension**. `/chrome` → **Enabled by default** avoids the
flag on future sessions at the cost of always-loaded browser tools.
3. **Grant site permissions** — these live in the extension's own settings (extension
icon → **Settings** → **Permissions**), not in Claude Code. Read-only calls run
without a prompt; state-changing calls always prompt and can be approved for the
whole site for the session. Granted permissions persist across sessions.
4. **Verify** — `/chrome` should report `Status: Enabled` / `Extension: Installed`.
Extension not detected, or commands stopping mid-run: dismiss any blocking browser
dialog, restart the extension in `chrome://extensions`, then `/chrome` →
**Reconnect extension**. "Receiving end does not exist" means the service worker went
idle — reconnect the same way. On Windows, named-pipe conflicts clear by restarting
Claude Code and closing other sessions using Chrome.
These are host instructions for the developer to run, never commands to execute on
their behalf. Extension site permissions are broader than the calling skill's scope:
keep to the origin that skill names regardless of what the extension allows.
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: 469k | 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: 321k | 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: 321k | 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: 321k | Updated | Claude CodeCodex | Verified | Install |