agent-loop-ext
Crash-resilient external agent loop with state persistence and CI/CD integration
Reconfigure the browser-control wiring. Rotate the Playwright MCP Bridge token, switch the target browser, or fully tear down the AIWG registration and provider injection. Walks the user through extension UI regeneration where needed; respects human-authorization gates for
$ npx -y skills add jmagly/aiwg --skill browser-reset --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/browser-resetContext preview
The summary Claude sees to decide when to auto-load this skill.
Reconfigure the browser-control wiring. Rotate the Playwright MCP Bridge token, switch the target browser, or fully tear down the AIWG registration and provider injection. Walks the user through extension UI regeneration where needed; respects human-authorization gates for
namespace: aiwg name: browser-reset description: Reconfigure the browser-control wiring. Rotate the Playwright MCP Bridge token, switch the target browser, or fully tear down the AIWG registration and provider injection. Walks the user through extension UI regeneration where needed; respects human-authorization gates for destructive operations. version: 0.1.0-draft status: draft platforms: [all] triggers: - "rotate browser token" - "reset browser control" - "change agent browser" - "tear down playwright mcp" - "switch from chrome to edge"
> Status: DRAFT — scaffolded from PoC. Implementation pending Inception outputs.
Recover from misconfiguration, rotate the per-install token, switch browsers, or fully tear down. Every destructive operation gated by `human-authorization`.
1. Show current setup (browser, token-file path, registry entry — token redacted) 2. Confirm intent (`AskUserQuestion`) 3. Walk user through extension regeneration:
4. Capture new token via interactive prompt 5. Atomic write to `~/.config/playwright-mcp/token` (mode preserved) 6. Update AIWG MCP registry env block: `aiwg mcp update playwright --env "PLAYWRIGHT_MCP_EXTENSION_TOKEN=<new>"` 7. Re-inject: `aiwg mcp inject --provider <p> --servers playwright` 8. Probe verify; report pass/fail
1. Show current browser 2. Detect alternatives (same logic as `browser-setup`) 3. Confirm switch (`AskUserQuestion`) 4. Walk user through extension install in new browser (if not already installed there) 5. Capture token from the new browser's extension 6. Update token file 7. Update AIWG MCP registry (no path changes, just env) 8. Re-inject 9. Probe verify
Note: this mode does **not** remove the old browser's extension or token. User can do that manually; we don't touch user's other browsers without explicit authorization.
1. Show what will be removed (registry entry, provider injection, token file path) 2. Explicit `human-authorization` gate (destructive) 3. `aiwg mcp remove playwright` 4. Provider config: remove `playwright` from `mcpServers` (we wrote it, we own removal) 5. Optionally delete `~/.config/playwright-mcp/token` (prompt) 6. Optionally delete `.aiwg/browser-allowlist.yaml` (prompt) 7. Note: AIWG never uninstalls the browser extension itself — that's a user action in the browser UI
For when the AIWG MCP registry is correct but the provider config was overwritten by another tool. Single command, no token capture.
1. Should "switch browser" preserve the old browser's wiring as a fallback? 2. Should "tear down" offer to also delete `~/.config/playwright-mcp/` dir if empty? 3. What's the canonical "reset to factory" command — call this skill, or a separate `aiwg run skill browser-uninstall`?
Reusable project context and specialist workflows for the AI tools you already use. Plan software, coordinate specialist reviews, prepare campaigns, investigate incidents, organize research, curate media, and maintain operational knowledge.
Repo: jmagly/aiwg
Crash-resilient external agent loop with state persistence and CI/CD integration
Detect requests for iterative autonomous agent loops and route to the appropriate loop executor
Automatically execute tests when code-generating agents modify source files, enforcing the execute-before-return pattern
Enable agent loops to learn from similar past tasks and share patterns across loops
Query and manage the executable feedback debug memory
Execute tests on generated code and iterate until passing