feat(team): add 'team auto-create' and 'team apply --agent' #237
Merged
Conversation
team auto-create <name>: bootstrap a two-agent claude+codex team in one shot — creates <name>-claude (agent-claude) and <name>-codex (agent-codex) envs, creates the team, and enrolls both with their runtimes. team apply --agent <id>: pick a team agent's latest submission directly, skipping verify/finalize. As an explicit human override it bypasses the verifier-verdict gate (no --force needed). Conflicts with --winner.
Extract the apply-by-agent and auto-create-roster logic out of the CLI handlers into testable team.rs core fns (apply_agent, latest_submission_for, auto_create_roster), and add unit tests: - apply_agent applies an agent's latest submission with no verify/finalize (verdict is None) and lands it on HEAD - apply_agent errors for an unknown agent and for an agent that hasn't submitted yet - auto_create_roster derives per-team env slugs (claude/codex) so teams never collide on env names
MANUAL.md: add the auto-create command row, extend the apply row with --agent (skip verify/finalize via explicit agent pick), and a shortcut block in the worked example. man/h5i.1: matching entries for team auto-create and team apply --agent.
Koukyosyumei
changed the title
Collaborator
Author
🪙 50% AI-authored · 7 files · prompt 40/100
Important 🎯 Goal: implement h5i team auto-create that creates claude+codex agent envs and wires them into a new team
💬 The ask
📍 What shipped
Note 🪴 Prompt maturity: 40/100 · developing · 5 prompts scored (83% of AI commits) 📊 heuristic breakdown
Readability (display-only): Flesch 73 · FK grade 7.5 · Gunning Fog 10.3 Offline classical-NLP score — no LLM. Higher = more specific, bounded, and grounded prompting. Tip ✅ Security scan cleanNo credentials leaked in this branch's diff. h5i scanned every added line against the secret rule pack and found nothing to rotate. Note Duplicate-code scan clean — no copy-paste blocks introduced. 🧠 Reasoning by file — 10633 nodes across 6 lanes, latest 24 only%%{init: {"flowchart": {"nodeSpacing": 42, "rankSpacing": 48, "diagramPadding": 14}, "themeVariables": {"fontSize": "18px"}} }%%
flowchart TB
subgraph lane_0["💭 Reasoning · 5 steps"]
direction LR
n_lane_0_0["THINK · Let me examine `discuss()` fully and how influence/independence tracking…"]:::t
n_lane_0_1["THINK · Let me find the enclosing function's signature to confirm `h5i_root` is …"]:::t
n_lane_0_2["THINK · both new ones). Let me run the env ingest tests (the changed call site) …"]:::t
n_lane_0_3["NOTE · DEFERRAL (~/Dev/h5i/man/man1/h5i.1): rence* the review rather than copy …"]:::n
n_lane_0_4["NOTE · DEFERRAL (~/Dev/h5i/src/team.rs): ecord the workflow lesson so I run the…"]:::n
n_lane_0_0 --> n_lane_0_1
n_lane_0_1 --> n_lane_0_2
n_lane_0_2 --> n_lane_0_3
n_lane_0_3 --> n_lane_0_4
end
subgraph lane_1["📄 src/team.rs · 5 ops"]
direction LR
n_lane_1_0["EDIT × 2"]:::a
n_lane_1_1["READ"]:::o
n_lane_1_2["EDIT × 2"]:::a
n_lane_1_0 --> n_lane_1_1
n_lane_1_1 --> n_lane_1_2
end
subgraph lane_2["📄 src/main.rs · 5 ops"]
direction LR
n_lane_2_0["READ"]:::o
n_lane_2_1["EDIT"]:::a
n_lane_2_2["READ × 3"]:::o
n_lane_2_0 --> n_lane_2_1
n_lane_2_1 --> n_lane_2_2
end
subgraph lane_3["📄 MANUAL.md · 3 ops"]
direction LR
n_lane_3_0["EDIT"]:::a
n_lane_3_1["READ"]:::o
n_lane_3_2["EDIT"]:::a
n_lane_3_0 --> n_lane_3_1
n_lane_3_1 --> n_lane_3_2
end
subgraph lane_4["📄 man/man1/h5i.1 · 2 ops"]
direction LR
n_lane_4_0["READ"]:::o
n_lane_4_1["EDIT"]:::a
n_lane_4_0 --> n_lane_4_1
end
subgraph lane_5["🗂 Other · 4 nodes"]
direction LR
n_lane_5_0["READ · read MEMORY.md"]:::o
n_lane_5_1["EDIT · edited MEMORY.md"]:::a
n_lane_5_2["EDIT · edited src/env.rs"]:::a
n_lane_5_3["EDIT · wrote clippy-all-targets-for-tests.md"]:::a
n_lane_5_0 --> n_lane_5_1
n_lane_5_1 --> n_lane_5_2
n_lane_5_2 --> n_lane_5_3
end
classDef o fill:#dbeafe,stroke:#1e3a8a,color:#0b1c4a;
classDef t fill:#fef3c7,stroke:#92400e,color:#3f2d05;
classDef a fill:#dcfce7,stroke:#166534,color:#0a2e16;
classDef n fill:#ede9fe,stroke:#5b21b6,color:#221251;
classDef m fill:#e5e7eb,stroke:#374151,color:#0b0f17;
Note 🪙 Token reduction — 8 captured tool outputs kept out of context: 17049 → 7064 tokens (58% saved, 9985 tokens). Full output is recoverable with 📜 Per-commit provenance — 6 AI-authored commits
|
Manual 'team add-env' gives each member a generated persona id (vega, hana, …) distinct from its runtime; auto-create was hardcoding the roster agent_id to 'claude'/'codex', conflating persona with runtime. Now it generates persona keys via gen_agent_id() (runtime still recorded separately) and prints the persona → runtime → env mapping so the keys are discoverable for 'team apply --agent <persona>'. - team.rs: drop agent_key from AutoMember (the env-slug/profile/runtime plan stays pure); generate the id in the handler, accumulating assigned ids so the two never clash - tests: roster test no longer asserts a fixed key; add a test that generated keys are distinct and never a runtime label - docs: MANUAL.md + man page note the generated persona ids and fix the 'apply --agent' example (was '--agent claude', now a persona key)
Previously 'team review submit' only recorded the review into the host-owned event log, so a confined target never received a peer's critique of its own work through its inbox (only the raw-git-read path exposed it). Now submit_review routes delivery through discuss(): post freeze it fans the review body into the target's per-env inbox AND records a discussion_msg, so the target's next revision is correctly marked non-independent (influenced by the review). Before freeze the review is still recorded for audit but not delivered, preserving the independence-first invariant (no cross-agent influence until first attempts are sealed). - submit_review gains an h5i_root param (needed for inbox fan-out); updated both host-side call sites (CLI + box-spool ingest) - tests: post-freeze delivery lands in the target inbox + marks influence; open-round review records but does not deliver - docs: MANUAL.md + man page note the new delivery + influence behavior
CI runs clippy with --all-targets --all-features; the test-only gen_agent_id(&[first.clone()]) tripped clippy::cloned_ref_to_slice_refs (my earlier --bin h5i clippy didn't cover test targets). Use std::slice::from_ref(&first).
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.
No description provided.