Worker factory must hand minted worker the original intent via factory_init skill

Mark commented

2026-04-10 19:40:37 +00:00

Owner

Architecture Violation

worker_factory provisions workers but never hands off the original intent. The comment at line 370 says "Intent will be reissued for claiming" but this never happens.

Current Flow (Broken)

mesh_node receives intent "write_article"
→ No workers for write_article
→ Falls back to worker_factory bidding
→ worker_factory wins bid
→ Provisions ephemeral_worker with CAPABILITIES=writer
→ Returns "Intent will be reissued for claiming"
→ Nothing happens - intent never reaches minted worker

Correct Flow

mesh_node receives intent "write_article"
→ No workers for write_article
→ Emits "provision_worker" intent with payload:
    - original_intent: "write_article"
    - target_capability: "writer"
→ worker_factory bids on provision_worker intent
→ worker_factory wins bid
→ worker_factory executes spawn_workers skill
→ Provisions ephemeral_worker with CAPABILITIES=writer
→ Uses factory_init skill to hand minted worker the original intent
→ Minted worker executes original intent directly
→ Minted worker emits result.committed

Required Fix

Implement factory_init skill in worker_factory:

After provisioning - Call factory_init skill
Direct handoff - Pass original intent to minted worker
Not backplane re-emission - Direct execution, not re-bidding
Minted worker executes - Worker runs, emits result

Implementation

# In worker_factory/run.py execute()

# After provisioning worker:
minted_worker_id = provisioning_result["worker_id"]

# Use factory_init skill to hand off intent
init_result = self.skills["factory_init"].execute(
    worker_id=minted_worker_id,
    original_intent=self.intent
)

# Minted worker now has intent, will execute
return {
    "ok": True,
    "provisioned_worker": minted_worker_id,
    "intent_handoff": "complete"
}

factory_init Skill Contract

skill: factory_init
input_schema:
  - worker_id: ID of minted worker
  - original_intent: Intent to hand off
output_schema:
  - handoff_status: "complete" or "failed"
  - execution_started: bool

Impact

Without this fix:

Worker spawning produces workers that never execute
Intents hang after provisioning
Factory pattern broken
Architecture violation propagates

References

capabilities/_muscle/worker_factory/run.py:370 - "Intent will be reissued" comment
docker/mesh_node.py - Fallback routing logic
AGENTS.md - Factory pattern section
docs/reference/FACTORY_PATTERN.md - Factory specification

## Architecture Violation worker_factory provisions workers but never hands off the original intent. The comment at line 370 says "Intent will be reissued for claiming" but this never happens. ## Current Flow (Broken) ``` mesh_node receives intent "write_article" → No workers for write_article → Falls back to worker_factory bidding → worker_factory wins bid → Provisions ephemeral_worker with CAPABILITIES=writer → Returns "Intent will be reissued for claiming" → Nothing happens - intent never reaches minted worker ``` ## Correct Flow ``` mesh_node receives intent "write_article" → No workers for write_article → Emits "provision_worker" intent with payload: - original_intent: "write_article" - target_capability: "writer" → worker_factory bids on provision_worker intent → worker_factory wins bid → worker_factory executes spawn_workers skill → Provisions ephemeral_worker with CAPABILITIES=writer → Uses factory_init skill to hand minted worker the original intent → Minted worker executes original intent directly → Minted worker emits result.committed ``` ## Required Fix Implement factory_init skill in worker_factory: 1. **After provisioning** - Call factory_init skill 2. **Direct handoff** - Pass original intent to minted worker 3. **Not backplane re-emission** - Direct execution, not re-bidding 4. **Minted worker executes** - Worker runs, emits result ## Implementation ```python # In worker_factory/run.py execute() # After provisioning worker: minted_worker_id = provisioning_result["worker_id"] # Use factory_init skill to hand off intent init_result = self.skills["factory_init"].execute( worker_id=minted_worker_id, original_intent=self.intent ) # Minted worker now has intent, will execute return { "ok": True, "provisioned_worker": minted_worker_id, "intent_handoff": "complete" } ``` ## factory_init Skill Contract ```yaml skill: factory_init input_schema: - worker_id: ID of minted worker - original_intent: Intent to hand off output_schema: - handoff_status: "complete" or "failed" - execution_started: bool ``` ## Impact Without this fix: - Worker spawning produces workers that never execute - Intents hang after provisioning - Factory pattern broken - Architecture violation propagates ## References - capabilities/_muscle/worker_factory/run.py:370 - "Intent will be reissued" comment - docker/mesh_node.py - Fallback routing logic - AGENTS.md - Factory pattern section - docs/reference/FACTORY_PATTERN.md - Factory specification

Mark referenced this issue

2026-04-10 22:04:11 +00:00

Terminology: template.yaml files are CONTRACT templates, not capability templates #117

Mark referenced this issue from a commit

2026-04-19 16:33:53 +00:00

Architecture fixes: worker handoff, starvation signals, bootstrap recovery

Mark commented

2026-05-05 03:41:46 +00:00

Author

Owner

🟡 PARTIALLY ADDRESSED — M009/S02: Capability Resolution in EphemeralWorker

Commit 705635f fixes one piece of the handoff: ephemeral_worker._execute_intent() now resolves intent-type capabilities (e.g. writing_task) to primitive IDs (e.g. writer_contract) before calling self.loader.get().

What was fixed

_execute_intent() calls resolve_primitive_for_intent(capability) before loading the contract
Prevents "Contract not loaded" errors when the factory hands off an intent-type instead of a primitive ID

What remains

The full factory_init skill contract (as described in this issue) is not yet implemented
After S02, the minted worker can load the right contract, but the original intent isn't directly handed off via factory_init
Comment at worker_factory/run.py line 370 ("Intent will be reissued for claiming") still reflects the current broken behavior

## 🟡 PARTIALLY ADDRESSED — M009/S02: Capability Resolution in EphemeralWorker Commit 705635f fixes one piece of the handoff: `ephemeral_worker._execute_intent()` now resolves intent-type capabilities (e.g. `writing_task`) to primitive IDs (e.g. `writer_contract`) before calling `self.loader.get()`. ### What was fixed - `_execute_intent()` calls `resolve_primitive_for_intent(capability)` before loading the contract - Prevents "Contract not loaded" errors when the factory hands off an intent-type instead of a primitive ID ### What remains - The full `factory_init` skill contract (as described in this issue) is not yet implemented - After S02, the minted worker can *load* the right contract, but the original intent isn't directly handed off via `factory_init` - Comment at worker_factory/run.py line 370 (`"Intent will be reissued for claiming"`) still reflects the current broken behavior ### Next - **S03**: Writer execution produces actual LLM output (needs vLLM integration) - **S04**: Full E2E test with real Redis/ledger/vLLM

Mark added the

feature-request

label

2026-05-12 23:26:56 +00:00

Mark added

blocked:needs-decision

and removed

feature-request

labels

2026-05-21 02:33:45 +00:00

Mark removed the

blocked:needs-decision

label

2026-05-28 07:19:13 +00:00

Mark added the

blocked:needs-decision

label

2026-06-07 05:47:57 +00:00

Mark commented

2026-06-07 05:47:57 +00:00

Author

Owner

⏸ Needs Decision

What needs to be decided?

Is this issue still relevant? If yes, remove the blocked:needs-decision label to route back to triage.

Blocking factors:

Didn't meet auto-approve criteria — scope=small, risk=low, spec=partial

Scope: small Risk: low Spec: partial Type: bug
Summary: Worker factory provisions workers via provision_worker but the original intent (e.g. 'write_article') never reaches the minted worker — it's returned in the factory response but nobody acts on it.
Suggestion: Add a factory_init internal function/skill in worker_factory that, after registering a new worker, directly spawns execution of the original_intent on the minted worker's capability channel — reuses existing WorkerSpawner or ephemeral worker path instead of re-emitting via backplane. The original_intent already flows through BidManager claim payloads and worker_factory responses; the handoff action is what's missing.
Rationale: Scope=small: only worker_factory/run.py needs the new skill logic and the provision_worker_v1 orchestration needs to invoke it after factory returns. Risk=low: data already flows (original_intent field in claim payloads, line 407 of worker_factory/run.py), the fix is additive and doesn't alter existing provisioning. Spec=partial: AGENTS.md mandates workers execute intents and factories define workers only, which this upholds, but factory_init as a named skill isn't documented. Type=bug: the current flow dead-ends after provisioning — the comment 'Handoff intent to spawned worker' at line 408/409 of run.py confirms this was intended but never implemented.

To unblock: Remove the blocked:needs-decision label. The triage will apply the appropriate outcome based on your choice above.

## ⏸ Needs Decision **What needs to be decided?** - **Is this issue still relevant?** If yes, remove the `blocked:needs-decision` label to route back to triage. **Blocking factors:** - **Didn't meet auto-approve criteria** — scope=small, risk=low, spec=partial --- **Scope:** small **Risk:** low **Spec:** partial **Type:** bug **Summary:** Worker factory provisions workers via provision_worker but the original intent (e.g. 'write_article') never reaches the minted worker — it's returned in the factory response but nobody acts on it. **Suggestion:** Add a factory_init internal function/skill in worker_factory that, after registering a new worker, directly spawns execution of the original_intent on the minted worker's capability channel — reuses existing WorkerSpawner or ephemeral worker path instead of re-emitting via backplane. The original_intent already flows through BidManager claim payloads and worker_factory responses; the handoff action is what's missing. **Rationale:** Scope=small: only worker_factory/run.py needs the new skill logic and the provision_worker_v1 orchestration needs to invoke it after factory returns. Risk=low: data already flows (original_intent field in claim payloads, line 407 of worker_factory/run.py), the fix is additive and doesn't alter existing provisioning. Spec=partial: AGENTS.md mandates workers execute intents and factories define workers only, which this upholds, but factory_init as a named skill isn't documented. Type=bug: the current flow dead-ends after provisioning — the comment 'Handoff intent to spawned worker' at line 408/409 of run.py confirms this was intended but never implemented. **To unblock:** Remove the `blocked:needs-decision` label. The triage will apply the appropriate outcome based on your choice above.

Mark added the

blocked:preflight

label

2026-07-31 04:26:46 +00:00

Mark commented

2026-08-11 04:32:37 +00:00

Author

Owner

Partially addressed by 705635f — removing blocked:needs-decision. Open if more work needed, close if the partial fix is enough.

Partially addressed by 705635f — removing blocked:needs-decision. Open if more work needed, close if the partial fix is enough.

Mark removed the

blocked:needs-decision