Skip to content

#370 Add complexity classification to assess-ticket and as standalone skill - #371

Merged
williamthorsen merged 1 commit into
mainfrom
370
Mar 27, 2026
Merged

#370 Add complexity classification to assess-ticket and as standalone skill#371
williamthorsen merged 1 commit into
mainfrom
370

Conversation

@williamthorsen

@williamthorsen williamthorsen commented Mar 27, 2026 •
edited
Loading

Copy link
Copy Markdown
Owner

Closes #370

What

Adds complexity as a fourth assessment dimension to assess-ticket and create a standalone /classify-complexity skill. Both use the existing 4-level rubric (trivial/mechanical/involved/architectural) from _data/complexity-classification.md but define their investigation processes independently — assess-ticket produces terse evidence bullets consistent with its other dimensions, while classify-complexity provides richer output with Scope, Drivers, and Risks sections.

Why

The complexity-classification rubric was only applied after planning or review by downstream consumer skills (wrap-up, next-steps-after-plan, next-steps-after-review). Knowing complexity before starting helps choose the right workflow upfront — whether to orchestrate, implement directly, or design first.

Details

Features

  • New classify-complexity skill with 5-step investigation process (resolve source, identify work surface, assess cross-cutting extent, assess decision density, classify) and structured output (Scope, Drivers, Risks)
  • Complexity dimension in assess-ticket: new investigation section, both output templates updated, complexity added to mode argument
  • Complexity uses a size scale (⚪/🟢/🟠/🔴) distinct from the concern scale (🟢/🟠/🔴) used by drift, relevance, and progress — documented in a separate emoji mapping subsection
  • Verdict reference table extended with a ⚪ column for the trivial level, using — for dimensions that don't have four verdicts

Test plan

  • Invoke /classify-complexity #370 and verify structured output with Scope, Drivers, Risks sections
  • Invoke /assess-ticket #370 (mode: all) and verify complexity appears as the fourth dimension
  • Invoke /assess-ticket #370 complexity and verify standalone complexity output
  • Verify verdict emojis match the size scale (⚪/🟢/🟠/🔴), not the concern scale

@williamthorsen
agents|feat: Add complexity classification to assess-ticket and as st…
24a1bb4 
…andalone skill

Add a fourth dimension (complexity) to assess-ticket alongside drift, relevance, and progress. Complexity uses a size scale (⚪/🟢/🟠/🔴) rather than the concern scale of the other dimensions, classifying tickets as trivial, mechanical, involved, or architectural using the existing 4-level rubric.

Create `/classify-complexity` as a standalone user-invocable skill with richer output (Scope, Drivers, Risks sections). Both skills reference the shared rubric in `_data/complexity-classification.md` but define their investigation processes independently.
@williamthorsen williamthorsen changed the title Add complexity classification to assess-ticket and as standalone skill #370 Add complexity classification to assess-ticket and as standalone skill Mar 27, 2026
@williamthorsen williamthorsen self-assigned this Mar 27, 2026
@williamthorsen williamthorsen added feature Added or improved external functionality scope:agents labels Mar 27, 2026
@williamthorsen
williamthorsen merged commit d3f2faa into main Mar 27, 2026
1 check passed
@williamthorsen
williamthorsen deleted the 370 branch March 27, 2026 14:32
williamthorsen added a commit that referenced this pull request May 4, 2026
@williamthorsen
release: agents-v0.2.0 run-core-v0.2.0 factory-v0.2.0 mcp-v0.2.0
17e4f03 
agents-v0.2.0
- #525 fix: Extract Jira-style ticket IDs from author-prefixed branches (#528)
- #452 feat: Have reviewers write findings incrementally for interruption resilience (#523)
- feat: Prevent agents from using interactive prompts
- #506 feat: Pre-load reviewer context for unfamiliar third-party APIs (#517)
- #511 internal: Migrate work-types to a JSON SSOT with schema validation (#516)
- #430 feat: Add preferences.yaml schema and normalize default_remote (#509)
- #497 feat: Show pros and cons in a list instead of inline (#503)
- #501 fix: Default merge-pr to remote-only branch deletion via tristate flag (#504)
- #494 feat: Resolve merge-pr scope and type via a tested script (#502)
- #491 feat: Canonicalize finding icons; switch Suggestion to ☝️ (#500)
- #496 feat: Mark skips and successes in install/uninstall output (#499)
- #462 feat: Add advisability dimension to /assess-ticket (#498)
- #382 feat: Add merge-pr skill family and rename merge config section (#495)
- #490 feat: Codify design priorities: correctness over convenience (#493)
- #489 feat: Reframe `## What` guidance around outside-reader audience (#492)
- #486 feat: Surface recommendation-gradient format at point of use (#488)
- #484 feat: Adopt plural tickets_created and drop counts (#487)
- #473 feat: Adopt cost-aware three-lane disposition for findings and follow-ups (#485)
- #476 feat: Add recommendation gradient to clarifying questions (#483)
- #478 feat: Persist deferred findings from /wrap-up sessions (#482)
- #474 refactor: Rename parse/resolve_prefix to title_format names (#481)
- #464 feat: Render ticket reference consistently in artifact headings (#479)
- #466 feat: Render commit, ticket, PR, and merge titles from declarative templates (#475)
- #471 fix: Make subagent guidance refs apply on Rovo Dev (#472)
- tooling: Exclude Playwright MCP files from linting
- #467 feat: Add update-jira-ticket skill to prevent INVALID_INPUT failures (#470)
- #461 feat: Append `Closes` line to PRs and expose `ticket_ref` (#465)
- #460 feat: Have /design-and-plan evaluate tickets on their merits (#463)
- #456 feat: Have Rovo Dev present choices as numbered text (#457)
- feat: Refine agent collaboration
- #446 feat: Require outcome-first framing in commit titles (#454)
- #448 fix: Have orchestrated-coder write change-summary incrementally (#453)
- #443 tooling: Automate replacement of dashed separator comments with headings or region folds (#451)
- #441 feat: Add provenance markers to generated files (#447)
- #442 fix: Prevent backtick over-escape in agent-authored GitHub bodies (#445)
- #437 fix: Eliminate relative Markdown links in installed guidance (#439)
- #426 feat: Make devlogs ticket-scoped and add linking frontmatter (#434)
- feat: If changes made, offer to run /create-pr after wrap-up
- feat: Improve question style in collaborative mode
- #416 feat: Add `generate label-map` CLI command with JSON Schema and readyup check (#420)
- docs: Record mistake writing long-running test
- #414 feat: Adopt release-notes voice in commit/change-summary skills (#417)
- feat: Instruct agents to follow best practices and prefer CLI tools
- #409 feat: Rationalize PR creation skills with platform-specific delegates (#411)
- #381 feat: Add GitHub label application to create-ticket skill (#410)
- #407 feat: Add project guidelines reading to 9 subagent definitions (#408)
- #384 feat: Improve outcome-first guidance across change-summary & commit skills (#406)
- #404 fix: Add documentation coverage convention to plan-producing skills (#405)
- feat: Improve guidance on writing What section of change summary
- fix: Prevent agents from categorizing trivial flaws as warnings
- #400 fix: Replace hardcoded artifact paths with placeholder in examples (#401)
- feat: Instruct agents to prefer git -C
- #388 fix: Filter stale entries from manifest on partial uninstall (#399)
- #386 tests: Add tests for describe-change.sh and installScripts (#398)
- #392 fix: List at-risk files in symlink safety error message (#397)
- #390 fix: Escalate test gaps for pipeline-authored code to F-level (#396)
- #391 fix: Resolve script paths at install time via template variable (#395)
- #393 fix: Add bin wrappers to eliminate pnpm install warnings (#394)
- #377 feat: Add guidance file install, uninstall, and status support (#389)
- docs: Add README
- #383 feat: Allow commit prefixes to be configured for user and for repo (#387)
- #379 feat: Require tests with code changes across orchestration pipeline (#380)
- #376 tooling: Migrate to nmr script runner (#378)
- #374 feat: Prompt for next steps after non-baseline assessment verdicts (#375)
- #372 feat: Skip complexity assessment when progress is complete (#373)
- fix: Ticket assessment checkboxes are not displayed in terminal
- #370 feat: Add complexity classification to assess-ticket and as standalone skill (#371)
- #135 feat: Add severity to legacy findings (#369)
- #363 fix: Rewrite relative Markdown paths to absolute during skill install (#368)
- #366 feat: Add PROJECT.md staleness check and agent launchers (#367)
- #364 feat: Add people-report skill (#365)
- #306 feat: Add shared complexity rubric and quick-fix pass to wrap-up (#362)
- fix: Gate next-steps presentation on reading reference file
- #360 fix: Replace plugin code-simplifier with standalone reviewer (#361)
- #357 feat: Add assess-ticket skill and extract shared ticket resolution (#359)
- #345 feat: Add staleness and relevancy check to design-and-plan skill (#356)
- #330 feat: Add numbered options and context-clearing to next-steps prompts (#355)
- #351 fix: Fix next-steps-after-plan over-recommending refinement (#354)
- #350 fix: Fix broken _data/ relative paths in skill files (#352)
- #348 feat: Add update-project-guidance skill (#349)
- #346 fix: Replace symlinks before writing generated files
- #346 fix: Install _data support files and filter dotfiles (#347)
- fix: Agents misunderstand session-context fallbacks
- #328 feat: Add ticket compliance checking to review-change skill (#331)
- #320 feat: Add next-steps resumption prompt to plan-producing skills (#329)
- #321 refactor: Remove stale get-session-context references from reviewers (#327)
- #319 refactor: Rename get-branch-context to get-session-context, centralize artifact resolution (#322)
- #315 feat: Add plan provenance to save-plan and refinedBy field (#317)
- #313 fix: Replace 24-hour active-run heuristic (#314)
- refactor: Clarify logic for finding agent preferences
- #290 feat: Show waiting-for-input state in factory visualization (#292)
- #282 fix: Clean up PR and review output conventions (#285)
- #281 feat: Add variable naming conventions (#284)
- refactor: Use short form of SHA in plan provenance metadata
- #280 fix: Remove ticket ID from condense-branch commit format
- #280 fix: Improve adherence to commit conventions (#283)
- #264 fix: Guard against zero parsed steps in high-trust plan conversion (#278)
- #266 feat: Add bb-pr-inline-comment skill (#276)
- feat: Add recommended next-step guidance to design-and-plan skill
- #267 docs: Document optional fields in artifact-conventions
- #267 feat: Wire up usage metrics for savings analysis (#275)
- #263 feat: Use plan provenance and trust level when calibrating orchestration effort (#265)
- fix: Prevent coder subagent from ignoring commit conventions
- tooling: Fix lint in package.json files
- fix: Coder ignores git-commit instructions on title content
- fix: Incorrect path to _data directory in skill definitions
- fix: PR body extracted from branch summary includes frontmatter
- #231 feat: Add design-and-plan skill for interactive design + planning (#234)
- #217 feat: Create find-orchestration-savings skill to identify token waste (#232)
- #105 feat: Add artifact-write safeguards to subagent prompts (#229)
- #223 feat: Create sprite-loading infrastructure for catwalk (#228)
- refactor: Discourage reliance on indirect sources
- #199 feat: Replace orchestration mode system with effort system (#206)
- #197 feat: Make review fan-out mode-aware and remove lite mode (#203)
- #169 feat: Replace timestamp prefixes with sequential counters on run artifacts (#170)
- #163 fix: Resolve artifact base directory from preferences instead of hardcoding project path (#168)
- #157 feat: Add /refine-plan skill for plan review and refinement (#159)
- #155 fix: Fix silent logging failure when MCP is unavailable to orchestrate engine (#158)
- #152 fix: Fix inconsistent artifact logging (#154)
- feat: Emit run_failed event when complete_run receives failed status (#150)
- #141 fix: Fix wrap-up skill action menu and summary narrative (#146)
- #121 fix: Restore platform-specific skill handling and prompts.yml generation (#144)
- #139 feat: Add --mode=lite to orchestrate-dev skill (#143)
- CODY-113 refactor: Deduplicate finding scheme from reviewer agents into shared skill (#134)
- CODY-87 feat: Migrate orchestrator to MCP and v3 events (#133)
- fix: Fix review-cycle doc gaps in Phase 4a entry and Phase 4b flow control
- fix: Address review findings for mode system and two-threshold model
- feat: Add mode system and two-threshold model to orchestrate-dev
- CODY-71 feat: Refine wrap-up: numbered findings, action menu
- CODY-71 feat: Add post-session wrap-up skill (#95)
- CODY-73 feat: Add rich summary sections to orchestrated run output (#88)
- CODY-65 feat: Update run-index.json incrementally during parallel review (#69)

factory-v0.2.0
- deps: Upgrade all deps to latest minor version
- #443 tooling: Automate replacement of dashed separator comments with headings or region folds (#451)
- #427 feat: Add `pick-demo-runs` script to rank archived runs (#436)
- #402 tooling: Add readyup and default kit (#403)
- deps: Upgrade all deps to latest minor version
- #376 tooling: Migrate to nmr script runner (#378)
- deps: Upgrade all deps to latest minor version & patch vuln
- deps: Upgrade all deps to latest minor version
- deps: Upgrade to Vite 8
- #76 feat: Enable playback of completed orchestrated runs (#344)
- #335 feat: Use slot-based positioning for orchestrator (#340)
- #311 feat: Animate office transitions (#336)
- #310 feat: Replace geometric placeholders with tileset rendering (#332)
- #318 tests: Require explicit method selection in silencedConsole (#326)
- #303 internal: Create 3-zone adapter for office visualization (#309)
- #299 feat: Shared logical scene state for visualizations (#304)
- #297 feat: Add ?format=html param to artifact content endpoint (#300)
- #293 spike: Design facility visualization with office prototype (#296)
- #290 feat: Show waiting-for-input state in factory visualization (#292)
- deps: Upgrade all deps to latest minor version
- #287 feat: Add three-zone factory-floor visualization (#291)
- #274 feat: Redesign orchestrator sprite (#286)
- #270 tests: Add edge-case tests for input deferral, playback, rebuild (#273)
- #259 feat: Add progressive artifact reveal to catwalk demo replay (#272)
- #260 feat: Expand block-robot sprites and improve palette contrast (#268)
- #239 fix: Fix catwalk layout geometry, actor rendering, and artifact positioning (#262)
- deps: Require Node 24+
- deps: Upgrade shared linting configs to latest version
- #247 feat: Generate HTML preview page for sprite sheets (#254)
- #236 fix: Subagent sprites are not centered on station x position (#250)
- #235 feat: Generate block-robot pixel art sprites (#249)
- #187 refactor: Move FlowDiagram to visualizations/flow/ (#244)
- tooling: Fix lint in package.json files
- #237 fix: Fix canvas not scaling up on viewport resize (#243)
- #238 fix: Guard against negative orchestrator station index (#242)
- #184 feat: Add chute animations, carried artifacts, and code badge (#241)
- #233 feat: Pulse scale instead of opacity; extract opacity constants (#240)
- refactor: Use preferences cascade for project base path
- #223 feat: Create sprite-loading infrastructure for catwalk (#228)
- #183 feat: Add animated state transitions to catwalk actors (#225)
- #213 feat: Add run directory scanning and refactor ProjectScanner (#221)
- #210 test: Add catwalk differ boundary tests for agent (#212)
- #182 feat: Implement catwalk config differ (#211)
- #207 fix: Skip interactive run directories in project scanner (#208)
- #199 feat: Replace orchestration mode system with effort system (#206)
- #181 feat: Integrate CatwalkScene with config-driven actor management (#201)
- #180 util: Implement static catwalk actors (#195)
- #179 util: Define CatwalkSceneConfig types and implement run-to-catwalk mapper (#192)
- #178 util: Implement catwalk layout engine (#191)
- #177 util: Add shared artifact color constants and catwalk dimension/timing constants (#190)
- #176 feat: Scaffold catwalk visualization with empty scene (#189)
- #171 feat: Gracefully handle invalid log files in orchestrated-run directories (#174)
- #148 feat: Propagate failure reason from run_failed event to CanonicalRunStatus (#167)
- tests: Add shared visualization mocks and status prop forwarding tests (#149)
- #10 tests: Silence console noise in test output (#147)
- #137 feat: Persist visualization mode in query string parameter (#142)
- #136 fix: Fix diagram view crash on undefined reviewers (#138)
- CODY-100 feat: Implement review-cycle visualization and iteration tracking (#129)
- CODY-116 fix: Fix canvas scaling to be width-driven with FitContainer (#125)
- CODY-99 feat: Add custom edge components with packet animation (#118)
- CODY-98 feat: Create custom node components for flow diagram (#112)
- CODY-97 feat: Add React Flow foundation and visualization switcher (#109)
- refactor: Migrate imports to @codeassembly/run-core (#104)
- CODY-94 fix: Handle newer parallelReview schema shapes in scene mapper (#102)
- CODY-79 feat: Show tooltip on hover over artifact boxes (#91)
- feat: Rebrand to "Code Assembly Factory"
- fix: Left-align labels, align upper reviewers, reposition artifacts
- fix: Fix orchestrator start position and approach distance
- CODY-63 feat: Add demo playback mode with event-sourced run replay (#81)
- CODY-60 feat: Render individual artifact boxes with configurable layout (#77)
- refactor: Fix lint surfaced by dependency upgrades
- CODY-67 feat: Move station labels to platform (#70)
- CODY-65 feat: Update run-index.json incrementally during parallel review (#69)
- CODY-59 feat: Improve representational quality of gates (#64)
- fix: Guard against stale setFacing from cancelled walk
- feat: Add directional sprite facing for approaching orchestrator
- CODY-42 feat: Position orchestrator to left of delegatee (#56)
- CODY-26 feat: Add resting animations for idle agents (#54)
- CODY-49 fix: Sync RunList selection with URL params and RunSelector dropdowns (#51)
- CODY-39 feat: Improve run listing readability in sidebar (#50)
- fix: Propagate null-safety to shared phase-inference module
- refactor: Simplify walk chain, deduplicate error handler, and clean up minor noise
- fix: Address review findings for artifact indicator
- feat: Add agent spawning, artifact indicator, and hand-off
- refactor: Simplify onDismissRun prop and clarify partial comment
- fix: Address review findings for settings persistence
- fix: Resolve lint errors in settings route test and hook
- feat: Update App and RunList for status-aware dismissal API
- feat: Rewrite useDismissedRuns hook with server persistence
- feat: Add fetchSettings and patchSettings client API functions
- feat: Register settings route and fix type compatibility
- feat: Add settings REST endpoints (GET/PATCH /api/settings)
- feat: Add SettingsStore service for persisting user settings
- feat: Add shared types and Zod schema for user settings
- fix: Position orchestrator at inferred current phase
- refactor: Consolidate phase-active checks to delegate to isPhasePresentInData
- feat: Show agents at stations before phase data is written
- CODY-41 fix: Correct structural element positioning in the factory scene (#44)
- MAC-35 feat: Add Zod schema validation for run-index.json (#43)
- CODY-34 fix: Accept null phases and criticality in status-adapter (#37)
- CODY-24 feat: Add orchestrator walk transitions along assembly line (#33)
- CODY-28 feat: Auto-refresh projects list with file watching (#32)
- refactor: Remove redundant ternary in camera bounds calculation
- feat: Add multi-level platforms and orchestrator positioning
- CODY-20 feat: Add RunList component with clickable run list in sidebar (#29)
- feat: Show project, ticket, run, status, duration in status  bar
- feat: Persist selected project, ticket, and run in URL
- CODY-7 feat: Add agent movement and status-driven animation transitions (#14)
- CODY-3 fix: Accept no-reason phase decisions in run-index validator (#13)
- fix: Load ImageSource data before engine start
- feat: Add sprite infrastructure and basic animations
- CODY-5 feat: Implement role-type architecture and all-phase agent mapping (#8)
- CODY-3 feat: Support run-index.json (v2) in status adapter (#4)
- fix: Skip missing status.json gracefully and extract shared isEnoent helper
- deps: Use exact version numbers in package.json
- refactor: Extract shared test fixtures and simplify buildGates
- tests: Add test coverage for game actors, FactoryScene, GameCanvas, and mappers
- fix: Fix StationActor graphics, GameCanvas race condition, and poll cleanup
- fix: Add onInitialize lifecycle hook to FactoryScene
- fix: Prevent isLoading flicker during poll cycles
- refactor: Add comment explaining type-narrowing aliases in useRunStatus
- tests: Strengthen error recovery assertions in useRunStatus test
- fix: Fix intervalRef leak on run switch, add missing test coverage
- tests: Add polling behavior tests for useRunStatus hook
- refactor: Remove dead typeof guards, redundant test assertions
- fix: Add route tests, fix ENOENT handling, and harden routes
- refactor: Simplify scanProject with early returns
- fix: Remove dead guard, improve project-scanner coverage
- fix: Fix duplicate tickets & stat check in project scanner
- fix: Reject null phase entries in status-adapter validation
- refactor: Simplify status-adapter validation helpers
- fix: Harden status-adapter type guard and validation

mcp-v0.2.0
- deps: Upgrade all deps to latest minor version
- #443 tooling: Automate replacement of dashed separator comments with headings or region folds (#451)
- deps: Upgrade all deps to latest version
- #376 tooling: Migrate to nmr script runner (#378)
- deps: Upgrade all deps to latest minor version
- tooling: Fix lint in package.json files
- fix: Exclude __tests__/ from staleness check to match compile scope
- refactor: Use preferences cascade for project base path
- #210 test: Add catwalk differ boundary tests for agent (#212)
- #199 feat: Replace orchestration mode system with effort system (#206)
- #163 fix: Resolve artifact base directory from preferences instead of hardcoding project path (#168)
- #160 feat: Warn when compiled output is stale if using development MCP code (#166)
- #152 fix: Fix inconsistent artifact logging (#154)
- #131: Verify end-to-end MCP integration (#151)
- feat: Emit run_failed event when complete_run receives failed status (#150)
- CODY-87 feat: Migrate orchestrator to MCP and v3 events (#133)

run-core-v0.2.0
- deps: Upgrade all deps to latest minor version
- #443 tooling: Automate replacement of dashed separator comments with headings or region folds (#451)
- #427 feat: Add `pick-demo-runs` script to rank archived runs (#436)
- #393 fix: Add bin wrappers to eliminate pnpm install warnings (#394)
- #376 tooling: Migrate to nmr script runner (#378)
- #76 feat: Enable playback of completed orchestrated runs (#344)
- #290 feat: Show waiting-for-input state in factory visualization (#292)
- tooling: Fix lint in package.json files
- fix: Add prepare script to fix bin linking in worktrees
- refactor: Use preferences cascade for project base path
- #217 feat: Create find-orchestration-savings skill to identify token waste (#232)
- #213 feat: Add run directory scanning and refactor ProjectScanner (#221)
- #199 feat: Replace orchestration mode system with effort system (#206)
- #171 feat: Gracefully handle invalid log files in orchestrated-run directories (#174)
- #148 feat: Propagate failure reason from run_failed event to CanonicalRunStatus (#167)
- #136 fix: Fix diagram view crash on undefined reviewers (#138)
- CODY-97 feat: Add React Flow foundation and visualization switcher (#109)
- refactor: Migrate imports to @codeassembly/run-core (#104)
- CODY-85 feat: Create @codeassembly/mcp server with run-data tools (#96)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

@williamthorsen williamthorsen

Labels

feature Added or improved external functionality scope:agents

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Add complexity classification to assess-ticket and as a standalone skill

1 participant

@williamthorsen