> Reusable prompt templates for each phase and task class.
Sign in to like and favorite skills
Reusable prompt templates for each phase and task class.
You are executing Phase 0 (Discovery Baseline) of specs/codex-claude-parity. Goal: - Build baseline capability inventory from .claude - Create outputs/P0_BASELINE.md - Create outputs/parity-capability-matrix.md Constraints: - Do not create .codex files - Do not modify .claude - Cite source paths for every claim Definition of done: - Both required outputs exist - Required capabilities are explicitly labeled - Risks and unknowns are documented - P1 handoff pair created
Inventory all relevant .claude artifacts by capability domain: - instruction/rules - skills - commands/workflows - context/handoff systems - quality/review patterns Output table fields: - capability - source artifact path - behavior summary - required/optional - candidate codex strategy
Identify parity risks before implementation. Classify each risk: - portability risk - ambiguity risk - validation risk - maintenance risk For each risk include severity, impact, mitigation, owner.
You are executing Phase 1 (Gap Analysis) of specs/codex-claude-parity. Input: - outputs/P0_BASELINE.md - outputs/parity-capability-matrix.md Output: - outputs/P1_GAP_ANALYSIS.md - outputs/parity-decision-log.md Task: - Map each required capability into one of: direct-port | adaptation | unsupported | defer - For every non-direct classification, provide rationale and mitigation. Constraint: - Analysis only. No .codex implementation in this phase.
Review parity-decision-log for weak entries. A valid entry must include: - capability name - classification - explicit decision - rationale - risk level - follow-up action Rewrite entries that are missing any field.
Produce a P2 readiness checklist from P1 outputs. Checklist sections: - required file scaffolding - instruction parity implementation tasks - workflow parity implementation tasks - validation hooks needed for P3
You are executing Phase 2 (Codex Config Implementation) of specs/codex-claude-parity. Input: - outputs/P1_GAP_ANALYSIS.md - outputs/parity-decision-log.md Deliver: - implement .codex/** according to approved mapping - create outputs/P2_IMPLEMENTATION_REPORT.md Constraints: - Keep .claude source assets intact unless explicitly required - preserve repo safety and workflow constraints - document every adaptation in the implementation report
Port high-value instruction constraints from .claude to .codex format. For each ported rule include: - source path - translated target path - any semantic drift - mitigation note
Map reusable .claude skill and command behaviors to Codex workflows. Output required in implementation report: - source workflow - codex equivalent - parity level (full/partial) - remaining gap
Evaluate deduplication using a symlink-first strategy. For each candidate path include: - source path - target path - portability assessment - fallback copy plan - validation step
Run a consistency pass: - verify .claude files are not unintentionally modified - verify .codex content does not contradict project safety rules - list unresolved implementation blockers
You are executing Phase 3 (Validation) of specs/codex-claude-parity. Deliver: - outputs/P3_VALIDATION_REPORT.md - outputs/parity-scorecard.md Required scenarios: - S1 spec bootstrap + handoff - S2 code edit + check workflow - S3 review workflow quality - S4 handoff + resume workflow - S5 symlink portability + fallback workflow Constraints: - provide command-level evidence - provide expected vs actual outcomes - score using RUBRICS.md worksheet
Audit validation report completeness. For each scenario verify: - objective - procedure - command evidence - observed results - pass/fail - follow-up Flag and repair incomplete scenarios.
Recompute weighted score from category scores. Check: - math consistency - acceptance gate thresholds - category minimum constraints If score below gate, identify exact blockers.
You are executing Phase 4 (Hardening & Final Handoff) of specs/codex-claude-parity. Deliver: - outputs/P4_HARDENING.md - updated REFLECTION_LOG.md - final handoff pair Tasks: - resolve residual defects from P3 - tighten docs and cross-links - record final lessons and follow-up plan Definition of done: - README required success criteria are met - rubric acceptance gates pass - no unresolved critical blockers
Audit all spec files for stale references, ambiguous wording, and inconsistent terminology. Fix: - phase naming mismatches - output filename drift - inconsistent definitions of parity
Prepare final handoff package for downstream maintenance. Must include: - current state summary - completed deliverables - unresolved items with owner/date - recommended next actions
Compare current phase outputs against README success criteria and MASTER_ORCHESTRATION gates. List: - criteria already met - criteria partially met - criteria unmet Recommend only highest-impact next actions.
Audit handoff files against context budget policy: - Working <= 2,000 tokens - Episodic <= 1,000 tokens - Semantic <= 500 tokens - Procedural links only - Total <= 4,000 tokens Report estimated token count and status.
Review parity-decision-log entries for weak rationale. A strong rationale includes: - technical compatibility reason - risk tradeoff - explicit mitigation Rewrite weak entries.
Use these blocks to compose task prompts.
Scope is restricted to specs/codex-claude-parity and .codex/.claude parity artifacts only.
Do not run destructive git commands or revert unrelated workspace changes.
Every non-trivial claim must include source path evidence.
At phase end, create both HANDOFF_P[N].md and P[N]_ORCHESTRATOR_PROMPT.md.
Produce all required files listed for this phase before claiming completion.
A high-quality phase prompt should:
| Defect | Impact | Fix |
|---|---|---|
| Missing output files | Incomplete phase artifacts | Add explicit output list |
| Vague "improve" wording | Low reproducibility | Replace with measurable tasks |
| No constraints | Scope drift | Add scope and safety clauses |
| No done criteria | Endless iteration | Add explicit completion gates |
| Mixed phases in one prompt | Confusion | Split into phase-specific prompts |
Execute Phase 1 (Gap Analysis) for specs/codex-claude-parity. Read: - specs/codex-claude-parity/outputs/P0_BASELINE.md - specs/codex-claude-parity/outputs/parity-capability-matrix.md - specs/codex-claude-parity/handoffs/HANDOFF_P1.md Produce: - specs/codex-claude-parity/outputs/P1_GAP_ANALYSIS.md - specs/codex-claude-parity/outputs/parity-decision-log.md Tasks: 1) classify each required capability as direct-port/adaptation/unsupported/defer 2) provide rationale and mitigation for non-direct classifications 3) define explicit prerequisites for P2 Constraints: - no .codex implementation changes in this phase - cite source paths for each classification Done when: - all required capabilities are classified - non-direct mappings include rationale and mitigation - P2 handoff pair is created
Use this skeleton for creating
P[N]_ORCHESTRATOR_PROMPT.md# P[N] Orchestrator Prompt: [Phase Name] Mission: Read first: Tasks: Outputs: Constraints: Definition of done:
Update this prompt library when: