synthesize
Skill15 starsChanged 15 days agoApache-2.0View on GitHub
Turn agent output into faithful, scannable, purpose-matched delivery. Use before writing or to restructure existing text. Routes to one need-specific shape without loading unrelated examples.
12 other projects use a version of this.
Adopted by
Most recently taken up by
- markusbegerow/board-of-directorstheir own version, first seen Oct 13, 2026
- Seanathon/vault-startertheir own version, first seen Sep 28, 2026
- dkpapadopoulos/auto-claude-skillstheir own version, first seen Sep 22, 2026
- iopsystems/skills-mcptheir own version, first seen Aug 24, 2026
- Kastalien-Research/thoughtboxtheir own version, first seen Jul 3, 2026
- patforna/core-skillstheir own version, first seen Jul 1, 2026
Other versions12 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 12
- Kastalien-Research/thoughtbox★ 64same name
- tuan3w/obsidian-vault-agent★ 39same name
- sparkline-ventures/product-eval★ 15same name
- dkpapadopoulos/auto-claude-skills★ 6same name
- iopsystems/skills-mcp★ 6same name
- markusbegerow/board-of-directors★ 5same name
- jasonmichaelbell78-creator/sonash-v0★ 2same name
- corygabrielsen/skills★ 1same name
- breethomas/bette-os★ 0same name
- Ekajto/my-claude-skills★ 0same name
2 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository95 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- agent-mailboxSkill
- agent-outputSkill
- ask-gateSkill
- ask-meSkill
- audit-coupled-constantSkill
- budget-checkSkill
- build-prototypeSkill
- caller-optionsSkill
- cold-reviewSkill
- debug-taskSkill
- define-termsSkill
- defuddleSkill
- design-an-interfaceSkill
- diff-recapSkill
- directives-auditSkill
- domain-checkSkill
- double-checkSkill
- extract-signalSkill
- freshness-contractSkill
- github-syncSkill
- handoff-executeSkill
- handoffSkill
- headless-claudeSkill
- headless-codexSkill
- headless-geminiSkill
- headless-grokSkill
- headless-modelsSkill
- headless-update-modelsSkill
- headless-usageSkill
- human-drawSkill
- human-mergeSkill
- human-outputSkill
- human-rewriteSkill
- implement-taskSkill
- improve-codebase-architectureSkill
- insight-adversarialSkill
- insight-anglesSkill
- insight-assessSkill
- insight-critiqueSkill
- insight-cross-examineSkill
- insight-exploreSkill
- insight-premortemSkill
- insight-reference-classSkill
- insight-retroSkill
- insight-root-causeSkill
- insight-skill-gapSkill
- insight-test-assumptionSkill
- insight-value-of-informationSkill
- investigateSkill
- json-canvasSkill
- map-checkSkill
- map-initSkill
- map-maintainSkill
- map-rulesSkill
- map-thisSkill
- monorepo-depsSkill
- multi-agent-modeSkill
- The first 60 shown.
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. 0 are common, 12 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 10 edited versions we read keep these.
12
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 36
Use it before writing, or to restructure existing text.
- Distinctive: line 43
Before composing, write a private numbered inventory of the protected units:
- Distinctive: line 46
every number with its unit.
- Distinctive: line 51
every condition that could reverse the conclusion
- Distinctive: line 53
Keep coupled units together: claim + evidence, attribution + stance, mechanism + duration, condition + consequence, risk + recourse, gate + …
- Distinctive: line 55
No protected unit may appear only below the fold.
- Distinctive: line 63
If a higher-authority instruction would require dropping, weakening, or changing a protected unit, stop and state the fidelity conflict.
- Distinctive: line 115
If no route matches, derive the minimum shape from first principles: what the reader must understand or do, the protected units, and the …
- Distinctive: line 126
Before a consequential action would be taken from the output, require a separate confirmation.
- Very specific: line 139
Use a plain table, list, or timeline directly when words and rows carry the relationship.
- Distinctive: line 145
Stop when another sentence changes neither understanding nor action.
- Distinctive: line 147
If the output is decision-bearing, close with the ordered next action, any applicable gate, and the condition that would change the …
- Distinctive: line 150
When the decision depends on the reader's values, frame viable options and the decision deadline.
- Very specific: line 158
If human-rewrite is installed, route repair there and retain the selected need.
“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/synthesize
Make reader-facing output cheap to understand without changing its meaning. Use it before writing, or to restructure existing text. If a simple answer is already minimal and clear, return it plainly. A no-op is valid.
Contract
Fidelity is the floor. Length is elastic. Every element needs a job.
Before composing, write a private numbered inventory of the protected units:
- negations, uncertainty, modality, scope, exceptions
- every number with its unit; both sides of any before/after pair
- attribution with stance; source disagreements with both sides attributed
- deadlines, defaults, required actions, ordered procedure steps
- irreversible actions, rejected options, method limits, minority evidence
- emotional or relational stakes when they affect the reader
- every condition that could reverse the conclusion
Keep coupled units together: claim + evidence, attribution + stance, mechanism + duration, condition + consequence, risk + recourse, gate + ask. No protected unit may appear only below the fold.
Restructuring never verifies. Output confidence cannot exceed source
confidence. Label every material claim confirmed, judgement, or estimate.
Unknowns are named gaps, not a fourth claim label. Name unchecked evidence when
it could change the decision. Do not add facts, settle ambiguity, or merge
conflicting sources silently.
If a higher-authority instruction would require dropping, weakening, or changing a protected unit, stop and state the fidelity conflict. Do not satisfy format or length authority by silently changing meaning.
Route and load
Route by what the reader must do. Load only the selected file.
| Reader need | Direct reference |
| Answer | answer |
| Get the point | point |
| Learn | learn |
| Investigate | investigate |
| Advice | advice |
| Decide | decide |
| Compare | compare |
| Prioritize | prioritize |
| Act now | act-now |
| Procedure | procedure |
| Plan | plan |
| Status | status |
| Monitor | monitor |
| What changed | change |
| Timeline | timeline |
| Organize | organize |
| Review | review |
| Feedback | feedback |
| Rewrite | rewrite |
| Explore | explore |
| Handover | handover |
| Record | record |
| Several needs | hybrid |
For substantive output, read grammar.md and the one selected need file completely. Hybrid additionally loads only the direct need file for each need represented in its modules, as instructed by hybrid.md. Confirm every required terminal marker in private scratch before composing. Do not expose the receipt unless a marker is missing or the reader asks. If a terminal marker is missing, reread that file once. If it remains missing, name the incomplete reference, apply this core, derive the minimum shape from the reader's need, and render word-only prose. Use the same fallback when a required reference is unavailable. Never imitate nearby formatting as a substitute.
Ambiguity rules:
- "What happened?" routes to Investigate while cause is open, Timeline when only settled event order matters.
- Reader wording wins when material suggests another need. Preserve the embedded decision or gap inside the requested shape.
- Hybrid is for genuinely separate needs, not an excuse for extra sections.
- If no route matches, derive the minimum shape from first principles: what the reader must understand or do, the protected units, and the shortest structure that preserves both. Do not use the compatibility index as a recipe.
Clarify and confirm
Clarify before composing only when an unresolved ambiguity would change the meaning, route, protected units, or requested action and cannot be preserved as a named gap. Ask the smallest question that settles it. Otherwise proceed and name the gap.
Before a consequential action would be taken from the output, require a separate confirmation. Read back the exact action, target, scope, material effect, and whether it can be undone. Synthesis may prepare that readback; it does not treat an earlier general request as confirmation for changed bytes, scope, risk, or external effects.
Compose
- Honor supplied authority, terminology, evidence, safety, and format rules.
- Inventory the protected units.
- Route and complete the required loads.
- Lead with the bottom line and place its confidence inside it. Add an early reversing condition only when one exists.
- Use a plain table, list, or timeline directly when words and rows carry the
relationship. If position, proportion, or connection needs a text figure,
conditionally load
human-draw; it owns everything inside the figure fence. - Render from the grammar and selected need. Examples teach shape only. Never reuse their facts, names, numbers, or judgments.
- Reconcile every inventory item and every spine element. Restore omissions. Stop when another sentence changes neither understanding nor action.
If the output is decision-bearing, close with the ordered next action, any applicable gate, and the condition that would change the recommendation.
When the decision depends on the reader's values, frame viable options and the decision deadline. Do not issue an imperative unless the reader supplied the criteria or requested your recommendation. State whose judgment a stance is and what would change it. Neutral records, procedures, explanations, and open exploration need no manufactured stance.
Existing text
If human-rewrite is installed, route repair there and retain the selected
need. Otherwise inventory, hoist the verdict and reversing caveats, apply the
selected need and grammar, cut only repetition or process narration, then
report the fidelity check in one line.
Several source reports require human-merge when installed. Without it, keep
each source attributed and separate, or ask which source governs.
Boundaries
human-output owns deep writing craft. human-rewrite repairs one existing
text. human-merge reconciles several reports. human-draw owns figures
inside fences; direct word, table, list, and timeline rendering stays here.
prose-audit diagnoses prose. agent-output governs agent-to-agent reports.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

