feat(zero-skill): async review + explicit token-vs-slug guidance#58
Open
rcliao wants to merge 2 commits into
Open
feat(zero-skill): async review + explicit token-vs-slug guidance#58rcliao wants to merge 2 commits into
rcliao wants to merge 2 commits into
Conversation
Two agent behavior improvements based on live interaction analysis: 1. **Review is now non-blocking**: reframes step 4 (Review) as async — agents deliver the result to the user first, then review. The end-to-end example now shows an explicit "deliver result" boundary before the review command. `zero runs --unreviewed` is the deferred review path. 2. **Token > slug for --capability**: agents were extracting `slug` from search results and passing it to `--capability` instead of the `token` field (confirmed via PostHog: 1.9.0 has 0 `from_search` fetches). Added explicit guidance that: - The token (`z_xxx.N`) is DB-backed and never expires (not session-scoped) - `--capability <token>` is required, not `--capability <slug>` - The `--capability` flag table and Gotchas bullet both reinforce this Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
spencercharest
added a commit
that referenced
this pull request
Jul 2, 2026
* feat(zero): add 15 agents/harnesses to the install matrix Folds the ZERO-294 adapter PRs (#37-#56, excl. #48/#58) into plugins/zero/agents.json: amp, antigravity, cline, continue-cli, copilot, crush, cursor, devin, goose, hermes, kilo-code, opencode, qwen-code, trae, zed. Roo Code (#53) and Replit Agent (#50) are not addable (product shut down / no local footprint). Existing kiro (#42) and warp (#46) entries were re-verified and stand as-is. qwen-code is plugin-installable (verified live: non-interactive marketplace install with --consent, native --auto-update); everything else is standalone-only. Hook fragments added where the host has a JSON hook config whose semantics tolerate our scripts: continue-cli (full Claude-compatible), devin (SessionStart+UserPromptSubmit), qwen-code (full, ms timeouts), copilot/cursor/goose (SessionStart only, host dialects). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * ci(version-check): detect manifests by top-level version key, not grep agents.json now carries host hook-dialect fragments that legitimately contain a nested "version" field (Copilot/Cursor hooks files); the grep matched it and lockstep failed on a non-manifest. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(zero): drop standalone.mcp from new matrix entries MCP config targets are only needed for ephemeral environments where auth session data cannot be written to disk; the new entries are all persistent hosts. kiro/warp keep their pre-existing blocks. The researched per-host MCP surfaces remain documented in the closed adapter PRs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(zero): drop standalone.mcp from kiro and warp too MCP config targets are only for ephemeral environments where auth session data cannot be written to disk; no matrix entry needs one. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
zero fetch, then review async (or deferred viazero runs --unreviewed). Previous wording implied blocking.--capability: live subagent test (2026-06-30) showed agents extractingslugfrom search results and using it in--capabilityinstead of thetokenfield — confirmed by PostHog (CLI 1.9.0 has 0from_searchfetches). Added explicit guidance at three points:--capability <token>bolded as "not the slug"--capabilityflag table + Gotchas bullet: both reinforcedContext
Subagent timing test showed Zero overhead = ~3.5s (search 2s + get 740ms + review 714ms). Making review async removes 714ms from the critical path that the user was waiting on. Token fix restores search attribution for 1.9.0 agents.
Test plan
z_xxx.Ntoken (not slug) in--capabilityafter searchzero review🤖 Generated with Claude Code