fix(spawn): unblock pipeline against migration 061 (G3 P0)

[namastex888]

Migration 061 (agents_id_shape_check + agent_templates UUID PK) went live without the spawn-side rewrites planned in wish G3. Result: every genie agent spawn errored with PostgresError and felipe's wish dispatches failed (autopg-distribution-cutover, agent-yaml-permissions-wireup).

Three minimum-blast-radius fixes against the cascade:

1. src/lib/agent-directory.ts:lookupTemplateTeam — WHERE id = $1 with a bare name (e.g. "engineer") tried to cast to UUID after 061 retyped agent_templates.id. Switched to WHERE name = $1 per the new schema.

2. src/lib/agent-registry.ts:saveTemplate — INSERT into agent_templates passed template.id (bare role name) into the now-UUID id column. Switched to inserting via the new name column with ON CONFLICT (name, team) WHERE name IS NOT NULL AND team IS NOT NULL so the partial unique index from 061 is the upsert target.

3. src/term-commands/agents.ts:registerSpawnWorker — set id to the durable identity UUID (ctx.agentIdentityId, populated from findOrCreateAgent) and stash the human-readable workerId in customName. Falls back to ctx.workerId only when agentIdentityId isn't populated (rare paths that 061 will reject loudly — exactly as designed). The register() ON CONFLICT (id) DO UPDATE clause merges runtime fields into the existing identity row instead of inserting a shadow.

Verification:

• bun --bun src/genie.ts agent spawn engineer --cwd /tmp --team genie --no-tui succeeds; agent show <name> shows UUID id; events errors --since 5m returns no patterns.

Wish: retire-session-names-id-only, Group 3 (P0).

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 2845a5cc-34af-4b33-96bb-def2356a1e96

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/spawn-pipeline-061-cascade

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[gemini-code-assist]

Code Review

This pull request updates the agent directory and registry to accommodate a database migration that changed the agent_templates.id field to a UUID and introduced a separate name column. Feedback highlights a logic error in saveTemplate where UUIDs might be incorrectly saved into the name column, potentially breaking lookups. Additionally, it is suggested to use ctx.agentName instead of ctx.workerId in registerSpawnWorker to ensure human-readable identifiers are preserved in the UI.

[gemini-code-assist]

The saveTemplate function uses template.id as the value for the name column in the agent_templates table. However, following migration 061, agent_templates.id is a UUID, and the rowToTemplate function (line 202) maps this database UUID to WorkerTemplate.id.

This creates a significant issue: any template read from the database and subsequently saved will have its UUID written into the name column. This breaks lookups that expect a human-readable name (such as lookupTemplateTeam in agent-directory.ts) and will result in duplicate template entries keyed by UUID instead of the role name. The WorkerTemplate interface and rowToTemplate should be updated to handle the new name column explicitly, and saveTemplate should use that field.

[gemini-code-assist]

In registerSpawnWorker, customName is set to ctx.workerId, and role falls back to ctx.workerId. Since generateWorkerId now returns a UUID (to satisfy the agents.id constraint), using it for customName results in non-human-readable identifiers in the UI (e.g., in genie ls) for legacy or explicit-role spawn paths.

It is better to use ctx.agentName, which contains the intended human-readable identity (e.g., "engineer" or the user-provided --role). This also ensures that the role field remains a valid template name for lookups if ctx.validated.role is missing.

    role: ctx.validated.role ?? ctx.agentName,
    skill: ctx.validated.skill,
    team: ctx.validated.team,
    customName: ctx.agentName,

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 563b3c3fd9

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve human template name in saveTemplate upserts

This upsert writes template.id into agent_templates.name, but template.id is sourced from listTemplates() where rowToTemplate maps it from the UUID agent_templates.id column. In call paths that round-trip templates (for example protocol-router reading a template, then calling saveTemplate to refresh lastSpawnedAt), this will insert/update rows under UUID-as-name instead of the agent name (e.g. engineer), causing template-key drift and duplicate template rows per team. That breaks subsequent name-based lookups and template matching.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Disambiguate template team lookup across teams

After this change, templates are upserted on (name, team), so multiple rows can legitimately share the same name across teams. This query still filters only by name with LIMIT 1, so it can return another team's row nondeterministically and pin spawns to the wrong team when the same role exists in multiple teams. The lookup needs an additional discriminator (or deterministic precedence) to avoid cross-team misrouting.

Useful? React with 👍 / 👎.