一键导入
dillylang-synthesize
Synthesizes upstream artifacts into an integrated proposal with explicit tradeoffs. Trigger= /dillylang-synthesize ARTIFACTS
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Synthesizes upstream artifacts into an integrated proposal with explicit tradeoffs. Trigger= /dillylang-synthesize ARTIFACTS
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Trigger: /compress-file FILEPATH or "compress <file>"
Maps problem structure to other domains, importing mechanism not metaphor. Trigger= /dillylang-analogize PROBLEM
Promotes exploratory prose into canonical docs through a conservative 5-stage gate — extract, route, judge, rank, synthesize. Model C recipe with operator-confirmed routing. Trigger= /dillylang-canonize PROSE
Decomposes a problem into axioms, derivations, and assumptions. Trigger= /dillylang-decompose PROBLEM
Judges an artifact against an explicit criterion. Trigger= /dillylang-evaluate CRITERION ARTIFACT
Applies Munger / Jacobi inversion on a problem statement. Trigger= /dillylang-invert PROBLEM
| name | dillylang-synthesize |
| description | Synthesizes upstream artifacts into an integrated proposal with explicit tradeoffs. Trigger= /dillylang-synthesize ARTIFACTS |
synthesize[[THIS is_grounded_by: urn:unique_reference:dillylang::spec-primer]]
Synthesize the given upstream artifacts into a single integrated proposal.
This is the rebuild step — not a summary, not a weighted average. If the result feels like averaging, you have not synthesized. Try again.
Calibration example:
Rejected: "Incorporates: invert output — provided useful failure analysis." (No specific contribution. What from the invert changed the proposal?)
Accepted: "Incorporates: invert_1 — FM-2 (schema split-brain) revealed that showing JSON and asking for prose causes the LLM to drop subfields. This drove the decision to use labeled prose templates instead of JSON blocks."
The integrated position. Must include:
What each upstream artifact contributed to the proposal. Every artifact must appear here or in open questions — no silent drops.
Each entry must include:
Each entry must include:
Where upstream artifacts disagreed or created tension.
Each entry must include:
What remains unresolved. Include deferred upstream artifacts here.
low | medium | high — with one sentence justifying the level.
n is sequential starting from 1 in each section.