ea2585e01a
chore: update GPT-5.2 references to GPT-5.4
...
Align runtime defaults, tests, docs, and generated artifacts with the newer GPT-5.4 baseline. Keep think-mode and prompt-routing expectations consistent after the model version bump.
2026-03-07 05:46:05 +09:00
bc82a5f28d
refactor(prometheus): align GPT prompt with 5.4 system prompt style
...
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode )
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai >
2026-03-06 13:44:20 +09:00
4b043c9d0b
refactor(agents/prometheus): remove unused barrel exports
...
Clean up unused re-exports from prometheus agents barrel file
Remove 9 unused exports identified by knip analysis
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode )
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai >
2026-02-23 02:43:16 +09:00
e87579b687
feat(agents): wire Gemini prompt routing into Sisyphus-Junior, Atlas, Prometheus
...
Add 'gemini' to prompt source types and route Gemini models to new
Gemini-optimized prompts via isGeminiModel detection. Update barrel
exports for all 3 agent modules. All existing tests pass.
2026-02-22 15:30:40 +09:00
0425c75447
feat(agents): add Gemini-optimized prompts for Sisyphus, Sisyphus-Junior, Prometheus, Atlas
...
Gemini models are aggressively optimistic and avoid tool calls in favor of
internal reasoning. These prompts counter that with:
- TOOL_CALL_MANDATE sections forcing actual tool usage
- Anti-optimism checkpoints before claiming completion
- Stronger delegation enforcement (Gemini prefers doing work itself)
- Aggressive verification language (subagent results are 'EXTREMELY SUSPICIOUS')
- Mandatory thinking checkpoints in Prometheus (prevents jumping to conclusions)
- Scope discipline reminders (creativity → implementation quality, not scope creep)
2026-02-22 15:30:40 +09:00
fd67e6aff5
feat(prometheus): add GPT-5.2 optimized prompt with model-based routing
...
- Create gpt.ts with XML-tagged, principle-driven prompt (Codex plan mode style)
- Add getPrometheusPrompt() routing: GPT models → GPT prompt, others → default
- Promote gpt-5.2 (high) to #2 in prometheus fallback chain
- Follow Atlas GPT variant pattern (isGptModel detection)
2026-02-19 18:40:42 +09:00
80bb262066
refactor(agents): convert all markdown tables to bullet lists across 12 agent files
...
Tables in template literal prompts render poorly in some LLM contexts.
Replaced 43 table instances with equivalent bullet list format preserving
all information. Affected: hephaestus, atlas/default, atlas/prompt-section-builder,
sisyphus-junior/gpt, librarian, explore, metis, prometheus/behavioral-summary,
prometheus/identity-constraints, prometheus/interview-mode, prometheus/plan-generation,
prometheus/plan-template.
2026-02-18 21:10:21 +09:00
a838925e23
Merge pull request #1535 from acamq/feature/start-work-plan-name-clean
...
feat(prometheus): include plan name in /start-work guidance
2026-02-18 18:20:08 +09:00
0037ad7185
fix(prometheus): replace single atomic write with incremental Write+Edit protocol ( #1904 )
2026-02-17 18:02:45 +09:00
54075d74e4
refactor: compress plan template while recovering lost specificity guidelines
...
Reduce plan-template from 541 to 335 lines by removing redundant verbose
examples while recovering 3 lost context items: tool-type mapping table in
QA Policy, scenario specificity requirements (selectors/data/assertions/
timing/negative) in TODO template, and structured output format hints for
each Final Verification agent.
2026-02-16 15:46:00 +09:00
9ba40b4d4c
enhance: enforce mandatory per-task QA scenarios and add Final Verification Wave
...
Strengthen TODO template to make QA scenarios non-optional with explicit
rejection warning. Add Final Verification Wave with 4 parallel review
agents: oracle (plan compliance audit), unspecified-high (code quality),
unspecified-high (real manual QA), deep (scope fidelity check) — each
with detailed verification steps and structured output format.
2026-02-16 15:46:00 +09:00
2d6b38b27e
enhance: calibrate Prometheus plan granularity to 5-8 parallel tasks per wave
...
Add Maximum Parallelism Principle as a top-level constraint and replace
small-scale plan template examples (6 tasks, 3 waves) with production-scale
examples (24 tasks, 4 waves, max 7 concurrent) to steer the model toward
generating fine-grained, dependency-minimized plans by default.
2026-02-16 15:14:25 +09:00
d9b2dd3f08
refactor(prompts): enrich explore/librarian delegation examples with structured context handoff
...
Expand prompt structure comment to 4-field format (CONTEXT/GOAL/DOWNSTREAM/REQUEST).
Update all explore/librarian task() examples across Sisyphus, Hephaestus,
Prometheus interview-mode, and both ultrawork variants with richer context
including downstream usage, scope limits, and return format expectations.
2026-02-10 14:24:18 +09:00
83595df63f
merge: integrate origin/dev into modular-enforcement branch
...
Resolves all merge conflicts, preserving our split module structure
while integrating all dev changes:
- Custom agent summaries support (parseRegisteredAgentSummaries)
- Background notification queue (enqueueNotificationForParent)
- Atlas shared git-worktree module (collectGitDiffStats, formatFileChanges)
- Ralph-loop withTimeout + DEFAULT_API_TIMEOUT=5000
- Session recovery assistant_prefill_unsupported error type
- Atlas agentOverrides forwarding
- Config handler plan model demotion (buildPlanDemoteConfig)
- Delegate-task agentOverrides, promptSyncWithModelSuggestionRetry, variant
- LSP init timeout + stale init detection
- isPlanFamily function + task-continuation-enforcer hook
- Handoff command
2026-02-08 17:34:47 +09:00
75124d0a83
fix: revert load_skills default and enforce via prompts instead
...
Revert .default([]) on load_skills schema back to required, restore the runtime error for missing load_skills, and add explicit load_skills=[] to all task() examples in agent prompts that were missing it.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-opencode )
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai >
2026-02-08 16:31:02 +09:00
cea97f896b
refactor: wave 2 - split atlas, auto-update-checker, session-recovery, todo-enforcer, background-task hooks
...
- Extract atlas/ into 15 focused modules (hook, event handler, tool policies, types, etc.)
- Split auto-update-checker into checker/ and hook/ subdirectories with single-purpose files
- Decompose session-recovery into separate recovery strategy files per error type
- Extract todo-continuation-enforcer from monolith to directory with dedicated modules
- Split background-task/tools.ts into individual tool creator files
- Extract command-executor, tmux-utils into focused sub-modules
- Split config/schema.ts into domain-specific schema files
- Decompose cli/config-manager.ts into focused modules
- Rollback skill-mcp-manager, model-availability, index.ts splits that broke tests
- Fix all import path depths for moved files (../../ -> ../../../)
- Add explicit type annotations to resolve TS7006 implicit any errors
Typecheck: 0 errors
Tests: 2359 pass, 5 fail (all pre-existing)
2026-02-08 15:01:42 +09:00
b30a131690
refactor: migrate delegate_task to task tool with metadata fixes
...
- Rename delegate_task tool to task across codebase (100 files)
- Update model references: claude-opus-4-6 → 4-5, gpt-5.3-codex → 5.2-codex
- Add tool-metadata-store to restore metadata overwritten by fromPlugin()
- Add session ID polling for BackgroundManager task sessions
- Await async ctx.metadata() calls in tool executors
- Add ses_ prefix guard to getMessageDir for performance
- Harden BackgroundManager with idle deferral and error handling
- Fix duplicate task key in sisyphus-junior test object literals
- Fix unawaited showOutputToUser in ast_grep_replace
- Fix background=true → run_in_background=true in ultrawork prompt
- Fix duplicate task/task references in docs and comments
2026-02-06 21:35:30 +09:00
252a3884a5
feat(prometheus): include plan name in /start-work guidance
...
Update plan-generation.ts to guide users to run /start-work with plan name.
For example: /start-work fix-bug instead of just /start-work
This makes it clearer which plan the user wants to execute.
2026-02-05 18:50:30 -07:00
1de6b52307
fix(prometheus): enforce path constraints and atomic write protocol ( #1414 )
...
* fix(prometheus): enforce path constraints and atomic write protocol
- Add FORBIDDEN PATHS section blocking docs/, plan/, plans/ directories
- Add SINGLE ATOMIC WRITE protocol to prevent content loss from multiple writes
- Simplify PROMETHEUS_AGENTS array to single PROMETHEUS_AGENT string
* fix: reconcile Edit tool signature in interview-mode.ts with identity-constraints.ts
Identified by cubic: Edit tool usage was inconsistent between files.
- interview-mode.ts showed: Edit(path, content)
- identity-constraints.ts showed: Edit(path, oldString="...", newString="...")
Updated interview-mode.ts to use the correct Edit signature with oldString and newString parameters to match the actual tool API and prevent agent hallucination.
2026-02-03 12:11:52 +09:00
817fc208b8
fix(prompts): add missing run_in_background and load_skills params to examples
...
All delegate_task examples now include required parameters to prevent
model confusion about parameter omission.
Fixes #1403
2026-02-03 10:50:26 +09:00
465c0168a5
refactor(prometheus): replace binary verification with layered agent-executed QA
...
Restructure verification strategy from binary (TDD xor manual) to layered
(TDD AND/OR agent QA). Elevate zero-human-intervention as universal principle,
require per-scenario ultra-detailed QA format with named scenarios, negative
cases, and evidence capture. Remove ambiguous 'manual QA' terminology.
2026-02-02 14:18:01 +09:00
c843254a86
refactor(agents): improve explore/librarian prompt examples with 4-part context structure
...
Add CONTEXT + GOAL + QUESTION + REQUEST structure to agent delegation examples.
This guides users to provide richer context when invoking explore/librarian agents.
2026-02-01 17:56:27 +09:00
fc70dd2492
refactor: major codebase cleanup - BDD comments, file splitting, bug fixes ( #1350 )
...
* style(tests): normalize BDD comments from '// #given' to '// given'
- Replace 4,668 Python-style BDD comments across 107 test files
- Patterns changed: // #given -> // given, // #when -> // when, // #then -> // then
- Also handles no-space variants: //#given -> // given
* fix(rules-injector): prefer output.metadata.filePath over output.title
- Extract file path resolution to dedicated output-path.ts module
- Prefer metadata.filePath which contains actual file path
- Fall back to output.title only when metadata unavailable
- Fixes issue where rules weren't injected when tool output title was a label
* feat(slashcommand): add optional user_message parameter
- Add user_message optional parameter for command arguments
- Model can now call: command='publish' user_message='patch'
- Improves error messages with clearer format guidance
- Helps LLMs understand correct parameter usage
* feat(hooks): restore compaction-context-injector hook
- Restore hook deleted in cbbc7bd0justsisyphus@users.noreply.github.com >
2026-02-01 16:47:50 +09:00
YeonGyu-Kim