chore(signals): support specifying harness for implementation tasks - #69833
Conversation
|
Reviews (1): Last reviewed commit: "add implementation harness" | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
There was a problem hiding this comment.
Pull request overview
Adds support for selecting a specific agent harness (runtime adapter), model, and reasoning effort when auto-starting “implementation” tasks in the Signals product, by resolving per-team runtime settings and threading them into task creation.
Changes:
- Introduces an
implementationstep constant for runtime configuration lookup. - Resolves the per-team agent runtime for implementation tasks and passes
runtime_adapter,model, andreasoning_effortinto task creation. - Adds a test ensuring the resolved runtime fields are forwarded to
tasks_facade.create_and_run_task.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| products/signals/backend/auto_start.py | Resolves implementation-step agent runtime and forwards runtime/model/effort into task creation. |
| products/signals/backend/agent_runtime.py | Adds STEP_IMPLEMENTATION constant for runtime resolution by step. |
| products/signals/backend/test/test_auto_start.py | Adds coverage for forwarding resolved runtime fields into task creation calls. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Small, contained addition to the self-driving product's internal agent-runtime config, authored by the owning team, following an existing established pattern (same resolution logic already used for other steps) and covered by a new test; the unresolved bot comments (validation ordering, latency) are enhancement suggestions on a defensive, fail-to-default design, not confirmed correctness or safety bugs.
- Author wrote 0% of the modified lines and has 33 merged PRs in these paths (familiarity MODERATE).
- copilot-pull-request-reviewer[bot], greptile-apps[bot] reviewed the current head.
- Greptile (P1) notes an unvalidated runtime/model/effort combo could be forwarded to the agent server and cause a misconfigured or failed task run — pre-existing pattern reused here, not a new gap, but worth tightening.
- Greptile (P2) / Copilot flag that resolve_agent_runtime's network read happens before the no-op early-return checks, adding avoidable latency on cheap no-op paths.
Gate mechanics and policy version
| Gate | Result | |
|---|---|---|
| prerequisites | ✓ | all clear |
| deny-list | ✓ | no deny categories matched |
| size | ✓ | 8L, 2F substantive, 51L/3F incl. docs/generated/snapshots — within ceiling |
| tier | ✓ | T1-agent / T1b-small (51L, 3F, single-area, chore) |
| stamphog 2.0.0b3 | .stamphog/policy.yml @ f3a2393 · reviewed head 1784c48 |
Support specifying harness, model and reasoning effort for implementation tasks.