Feature: Automation launch integration

[PredictabilityAtScale]

Context

Delivers the guarded automation handoff described in plans/feature-automation-integration.md, enabling IssueTriage users to launch downstream AI agents responsibly.

Goals

• Pre-launch validation with readiness thresholds, metadata checklist, and override logging.
• Pluggable automation adapter interface starting with GitHub Copilot agent support.
• Status polling, notifications, and kill-switch handling within VS Code.
• Outcome recording that feeds back into assessment history and telemetry.

Acceptance Criteria

• IssueTriage: Launch Automation is enabled only for issues meeting the configured readiness threshold or authorized overrides.
• Launch flow packages assessment context, triggers the default adapter, and displays progress updates in the panel/status bar.
• Users can cancel in-flight runs, with state cleanup and user feedback.
• Completion records (success/failure, duration, quality notes) persist and surface in the dashboard timeline.
• Adapter interface (IAutomationAdapter) documented and tested to support future targets.

Dependencies

• Stable assessment outputs with confidence metrics.
• Credential storage for automation endpoints.
• Event bus for broadcasting state changes to UI components.

Notes

• Begin with a single adapter; design audit logging hooks for enterprise tiers even if stubbed for MVP.

[PredictabilityAtScale]

IssueTriage Assessment

Model: openai/gpt-5-mini
Issue: PredictabilityAtScale/IssueTriage #7
View issue

Dimension	Score
Composite	66.3
Requirements	65.0
Complexity	75.0
Security	55.0
Business Impact	70.0

Summary: The issue defines a clear high-level goal and acceptance checklist for a guarded automation launch flow, but lacks important implementation contracts, data formats, and test fixtures that an autonomous agent needs to implement it end-to-end. Key gaps include exact assessment output schema and confidence thresholds, the adapter interface signature and lifecycle, credential APIs, event-bus contracts, and concrete UI/state-machine behavior for cancellation and progress. Security and audit/logging requirements are noted but underspecified for enterprise needs. Providing small, machine-checkable specifications, mocks, and test cases would make this feature much more automatable.

Recommendations:

• Specify machine-readable contracts: provide TypeScript interfaces and JSON schema for (a) assessment outputs including confidence metrics and required metadata fields; (b) the IAutomationAdapter public API (methods, input/output shapes, error codes, lifecycle events); and (c) the event-bus message types for launch/poll/kill/completion so an agent can generate code and tests.
• Add concrete acceptance tests and test fixtures: include unit tests that assert readiness-threshold gating, an override flow (with logged override records), adapter invocation, progress updates, cancellation behavior, and completion recording. Provide mock adapter implementations and fixtures for assessment outputs at different confidence levels.
• Document credential and audit requirements: list storage mechanism (e.g., VS Code secrets API or encrypted storage), required scopes/permissions for target adapters (e.g., Copilot), and an audit logging interface (with examples of fields to record for enterprise tiers). Provide a safe default stub that records logs to local telemetry during development.
• Define UI/state-machine and cancel semantics: provide a small state diagram or step list for launch lifecycle (ready -> launching -> running -> cancelling -> completed/failed), polling cadence, timeout and kill-switch behavior, user feedback messages, and cleanup steps for cancelled runs so automation can implement robust cancellation and recovery.
• Provide CI-friendly integration tests and rollout/rollback plan: include an integration test that exercises a real or emulated adapter end-to-end (runable under CI with mocked credentials), plus migration/rollback steps for persisted outcome records (schema migrations), and clear metrics to mark success (e.g., duration, quality notes).

Last updated: 10/19/2025, 6:53:44 PM

[PredictabilityAtScale]

IssueTriage Assessment

Model: openai/gpt-5-mini
Issue: PredictabilityAtScale/IssueTriage #7
View issue

Dimension	Score
Composite	50.0
Requirements	65.0
Complexity	75.0
Security	55.0
Business Impact	70.0

Summary: The issue defines clear high-level goals and acceptance criteria for a guarded automation launch flow (gating, adapter interface, polling, cancellation, outcome recording). However, critical integration contracts are missing—most importantly the assessment output schema, the exact adapter target API, and the credential storage surface—preventing a deterministic v1 implementation. With those clarifications an autonomous agent can implement a working MVP and tests.

Pre-implementation questions:

• Provide a machine-readable assessment output schema (field names and the confidence metric name/range) so the agent can deterministically gate launches.
• Confirm the initial adapter target and invocation surface: e.g., VS Code Copilot extension API, Copilot service API, or another endpoint—specify which one should be used for v1.
• Specify the credential storage mechanism for v1 (e.g., VS Code Secrets API or a provided test/stub) and whether sandbox/test credentials or endpoints are available.

Last updated: 11/3/2025, 1:17:55 PM

[PredictabilityAtScale]

IssueTriage Assessment

Model: openai/gpt-5-mini
Issue: PredictabilityAtScale/IssueTriage #7
View issue

Dimension	Score
Composite	50.0
Requirements	65.0
Complexity	75.0
Security	55.0
Business Impact	70.0

Summary: The issue defines clear high-level goals and acceptance criteria for a guarded automation launch flow (gating, adapter interface, polling, cancellation, outcome recording), but key integration contracts are missing. Critical gaps—assessment output schema (confidence metric), the exact adapter target/API to call, and the credential-storage surface—prevent a deterministic v1 implementation. With those clarifications an autonomous agent can implement a working MVP and tests.

Pre-implementation questions:

• Which adapter target should v1 integrate with (e.g., GitHub Copilot service/API, the VS Code Copilot extension API, or a mocked/emulated adapter endpoint)?
• Provide a machine-readable assessment output contract: exact field names for readiness/confidence (name and numeric range), required metadata checklist fields, and the configured readiness threshold semantics.
• Specify the credential-storage mechanism for v1 (e.g., VS Code Secrets API, a test/stub credential store, or an external vault) and whether CI/test credentials/endpoints will be provided.

Last updated: 11/3/2025, 4:04:28 PM

[PredictabilityAtScale]

IssueTriage Assessment

Model: openai/gpt-5-mini
Issue: PredictabilityAtScale/IssueTriage #7
View issue

Dimension	Score
Composite	50.0
Requirements	65.0
Complexity	75.0
Security	55.0
Business Impact	70.0

Summary: The issue lays out clear high-level goals and acceptance criteria for a guarded automation launch flow (gating, adapter interface, polling, cancellation, outcome recording). However, critical integration contracts are missing that prevent an autonomous agent from producing a deterministic v1 implementation. The three primary gaps are the assessment output schema (readiness/confidence metric), the concrete v1 adapter target, and the credential-storage surface. Once those are specified an agent can implement the feature and tests.

Pre-implementation questions:

• Which adapter target should v1 integrate with: GitHub Copilot service API, the VS Code Copilot extension API, or a mocked/emulated adapter endpoint?
• Provide the machine-readable assessment output contract used for gating (exact readiness/confidence field name and numeric range, plus any required checklist fields).
• Specify the credential-storage mechanism for v1 (e.g., VS Code Secrets API, a provided test/stub, or an external vault) and whether test/CI credentials or endpoints will be supplied.

Last updated: 11/13/2025, 1:50:39 PM