Conversation
… and provenance Extract research-orchestrator agent from start skill to enable reuse across start, update, and augment commands. Add four blocking codex review gates (post-findings, post-merge, post-report, post-issues) with quarantine-on-failure semantics. Enforce source provenance on every finding with full URL re-verification during review. Key changes: - New agents/research-orchestrator.md owns all phase management - skills/start/SKILL.md gutted to thin launcher (-540 lines) - Dimension-analyst gains self-reflection protocol, WebSearch retry, and inline provenance schema - New skills/update/SKILL.md with swarm orchestration + delta detection - New skills/schedule/SKILL.md for Desktop Scheduled Task management - Lineage tracking in state.json for full research provenance chain - research-progress.md rendered on phase transitions (harness pattern) - /sigint:resume reads progress file first (harness init protocol) - Blackboard dual-write (blackboard + file) as default, not fallback - Fix trend-modeling blackboard key conflict (findings_trend_modeling) - Plugin version bump to 0.5.0
There was a problem hiding this comment.
Pull request overview
This PR restructures sigint’s research workflow around a reusable research-orchestrator harness that owns phase management, codex review gates, provenance requirements, and continuity artifacts (progress + lineage), while turning existing commands/skills into thin launchers.
Changes:
- Introduces
agents/research-orchestrator.mdimplementing the harness pattern, provenance schema, delta detection, and multiple codex review gates. - Refactors
/sigint:startand/sigint:updateto delegate orchestration to the new agent; adds scheduling skill/command for recurring updates. - Clarifies blackboard key separation for trend analysis vs trend modeling; updates docs/evals/plugin version to 0.5.0.
Reviewed changes
Copilot reviewed 15 out of 15 changed files in this pull request and generated 7 comments.
Show a summary per file
| File | Description |
|---|---|
agents/research-orchestrator.md |
New orchestrator agent defining phases, gates, provenance, delta detection, progress/lineage behaviors. |
skills/start/SKILL.md |
Simplified into a thin launcher that delegates to the orchestrator. |
skills/update/SKILL.md |
New update skill that loads prior state and delegates update-mode orchestration. |
skills/schedule/SKILL.md |
New scheduling skill for recurring update runs via Desktop Scheduled Tasks. |
commands/update.md |
Updated command to execute the update skill (pass-through arguments). |
commands/schedule.md |
New command shell delegating to the schedule skill. |
commands/resume.md |
Updated to prefer research-progress.md first (harness init protocol) and include lineage/quarantine context. |
agents/dimension-analyst.md |
Adds provenance requirements, retry protocol, self-reflection, and dual-write guidance. |
agents/report-synthesizer.md |
Adds a blocking post-report codex gate request flow. |
agents/issue-architect.md |
Adds a blocking post-issues codex gate request flow. |
skills/trend-analysis/SKILL.md |
Clarifies blackboard key usage. |
skills/trend-modeling/SKILL.md |
Fixes blackboard key to findings_trend_modeling. |
docs/reference/agents.md |
Documents orchestrator, tools, modes, and adds trend_modeling mapping. |
evals/agents/research-orchestrator/evals.json |
Updates eval to include trend_modeling as an eighth dimension. |
.claude-plugin/plugin.json |
Bumps plugin version to 0.5.0. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Address Codex adversarial review findings: 1. Update-mode merge now reconciles against prior state instead of blindly appending. UPDATED findings replace prior entries in-place, CONFIRMED findings get a last_confirmed timestamp, POTENTIALLY_REMOVED findings move to archived_findings[] instead of staying in the active array. Delta detection runs BEFORE merge to drive reconciliation. 2. Remove /sigint:schedule command and skill — no scheduler backend exists yet (Desktop Scheduled Task API is unspecified). Shipping a placeholder that silently pretends to succeed is worse than not shipping it. Will re-add when a concrete scheduler integration is available.
…, JSON, progress file - Fix --delta/--no-delta flag inconsistency: default enabled, --no-delta disables - Add --topic flag to update for explicit session selection - Make post-report and post-issues codex gates self-contained instead of blocking on team-lead SendMessage (eliminates deadlock risk) - Fix single-quoted JSON response templates to valid double-quoted JSON - Fix progress file: append status sections instead of regenerating (preserves audit trail from phase transition log entries)
Ran autoresearch improvement loops on all skills modified in this PR. Results: - start: 100% baseline (no changes needed) - update: 36% → 65% — clearer argument parsing, structured error paths showing planned workflow, explicit reconciliation semantics - trend-analysis: 94% → 100% — macro trend category labeling, emerging signal implication language, three-valued logic intro - trend-modeling: 83% → 98% — mandatory correlation notation, extended notation definition table, large model (7+ var) guidance Also adds evals for start (9 cases) and update (8 cases) skills that previously had none.
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
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.
Summary
New files (4)
Modified files (11)
Test plan