1---2name: coding-agents-prompting-flow-23description: Workflow for authoring and adapting AI-agent prompts: skills, agents, workflows, rules, etc.4---56<coding-agents-prompting-flow>78<description_and_purpose>9Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blueprint -> for_each_prompt_loop(draft -> hardening -> edit) -> simulate -> validate`. Every phase logs status in coding-agents-prompting-flow-state.md, Prompt Brief carries through all phases, and final output traces to request intent.10</description_and_purpose>1112<workflow_phases>13141. Orchestrator must trust the system and skills; coordinate only sequence, artifacts, state, and approvals.152. Load only references needed for the current phase.163. Execute phases sequentially, do not skip!1718<prerequisites>19201. All Rosetta prep steps MUST be FULLY completed212. USE SKILL `load-project-context`, `orchestration`, `hitl`223. MUST ALWAYS use todo tasks ledger, ASAP. Phases are sequential. Independent tasks can run in parallel.234. MUST just-in-time load/execute/update each phase's: instructions, definitions, skills, state file.245. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`.256. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping.267. This workflow MUST be used with Fable, Opus, GPT-5.5+ class models => IF NOT - DEMAND USER TO SWITCH MODEL.278. Workflow state MUST be saved to `agents/TEMP/<FEATURE>/coding-agents-prompting-flow-state` file.2829</prerequisites>3031<discover step="1" subagent="discoverer" role="Context discoverer" subagent_required_model="gpt-5.6-terra-medium">32331. Discover project-local context, relevant prompt-family artifacts, and required references for this request.342. Input: request + optional existing prompt. Output: `Discovery Notes` + `Reference Set`.353. Update `coding-agents-prompting-flow-state.md`.364. HITL when discovered context conflicts with user intent or critical references are missing.3738</discover>3940<extract_intake step="2" subagent="prompt-engineer" role="Intent extractor" subagent_required_model="gpt-5.6-sol-high">41421. Extract requirements from source prompt when present and intake clarifications from user.432. Input: request + optional existing prompt + `Discovery Notes` + `Reference Set`. Output: `Prompt Brief` + `Open Questions`.443. Update `coding-agents-prompting-flow-state.md`.454. HITL explicit approval required for `Prompt Brief`.4647</extract_intake>4849<blueprint step="3" subagent="prompt-engineer" role="Blueprint Architect" subagent_required_model="gpt-5.6-sol-high">50511. Design blueprint: structure, actors, contracts, and boundaries for target prompt set.522. Input: approved `Prompt Brief`. Output: `Blueprint`.533. Update `coding-agents-prompting-flow-state.md`.544. HITL approval when architecture or tradeoffs are ambiguous.5556</blueprint>5758<for_each_prompt_loop step="4" subagent="prompt-engineer" role="Prompt Author" subagent_required_model="gpt-5.6-sol-high">59601. Scope: `draft` target prompts. Subagents must draft one individual prompt file or surgical changes to one file at-a-time: it must think thoroughly about each file first, then it provides the full structure of the file with specific grounded points each section should contain or change, then it validates with schemas requirements/suggestions, and only then it can generate actual file. Repeat for each file. ALSO draft means fully ready for review prompt, just not reviewed and approved! Aim newly created files to have small extensive coverage, clear concerns, while each rules is very small612. Input: approved `Prompt Brief` + `Blueprint`. Output: `Draft Prompt Set` + optional change-log.md in FEATURE PLAN folder.623. Update `coding-agents-prompting-flow-state.md`.634. HITL when loop stalls, conflicts appear, or intent becomes unclear.6465</for_each_prompt_loop>6667<for_each_prompt_loop step="5" subagent="prompt-engineer" role="Prompt reviewer and hardening" subagent_required_model="gpt-5.6-sol-high">68691. Run loop for each target prompt: `hardening -> edit` until pass criteria or HITL decision.702. This is automated review by subagent, this is not HITL review!713. Input: approved `Prompt Brief` + `Blueprint` + `Draft Prompt Set` + prompt family information. Output: `Prompt Set` + optional change-log.md in FEATURE PLAN folder.724. Update `coding-agents-prompting-flow-state.md`.735. HITL when loop stalls, conflicts appear, or intent becomes unclear.7475</for_each_prompt_loop>7677<simulate step="6" subagent="prompt-engineer" role="Execution tracer" subagent_required_model="gpt-5.6-sol-high">78791. Simulate realistic runs and trace context/cognitive load across the prompt chain.802. Input: `Prompt Brief` + `Prompt Set`. Output: `Simulation Notes`.813. Update `coding-agents-prompting-flow-state.md`.824. HITL when simulation exposes major behavioral risk.8384</simulate>8586<validate step="7" subagent="prompt-engineer" role="Quality validator" subagent_required_model="gpt-5.6-sol-high">87881. Validate final artifacts against intent, contracts, failure modes, and traceability.892. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md).903. Update `coding-agents-prompting-flow-state.md` and mark complete.914. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message.9293</validate>9495</workflow_phases>9697<validation_checklist>9899- `discover` runs first and produces `Discovery Notes` + `Reference Set`.100- `Prompt Brief` is approved before blueprint starts.101- Adaptation requests load `pa-adapt.md` and keep source intent traceable through the target artifact.102- `Prompt Brief` is present as input in loop, simulation, and validation phases.103- Loop explicitly runs `draft -> hardening -> edit` for each target prompt.104- Every phase has artifact evidence and state update in coding-agents-prompting-flow-state.md in FEATURE TEMP folder.105- `Validation Pack` includes checklist results, tests, failure modes, and traceability.106- Workflow stays thin: no skill-internal templates or execution internals are embedded.107- Final outputs map to request intent with explicit traceability.108109</validation_checklist>110111<pitfalls>112113- Orchestrator executes skill internals instead of coordinating contracts.114- `discover` is skipped or delayed after intake.115- `Prompt Brief` is treated as optional in downstream phases.116- Loop is collapsed into one draft pass without hardening/edit.117- Too many references are loaded; phase-scoped loading is skipped.118- State file in FEATURE TEMP folder is stale or missing phase artifact evidence.119120</pitfalls>121122</coding-agents-prompting-flow>
griddynamics