# Send Decide to your agent Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually. ## Fast path - Download the package from Yavira. - Extract it into a folder your agent can access. - Paste one of the prompts below and point your agent at the extracted folder. ## Suggested prompts ### New install ```text I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete. ``` ### Upgrade existing ```text I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run. ``` ## Machine-readable fields ```json { "schemaVersion": "1.0", "item": { "slug": "decide", "name": "Decide", "source": "tencent", "type": "skill", "category": "AI 智能", "sourceUrl": "https://clawhub.ai/ivangdavila/decide", "canonicalUrl": "https://clawhub.ai/ivangdavila/decide", "targetPlatform": "OpenClaw" }, "install": { "downloadUrl": "/downloads/decide", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=decide", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "packageFormat": "ZIP package", "primaryDoc": "SKILL.md", "includedAssets": [ "SKILL.md", "components.md", "confidence.md", "exceptions.md", "memory-template.md", "migration.md" ], "downloadMode": "redirect", "sourceHealth": { "source": "tencent", "slug": "decide", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-05-01T01:10:06.270Z", "expiresAt": "2026-05-08T01:10:06.270Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=decide", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=decide", "contentDisposition": "attachment; filename=\"decide-1.0.1.zip\"", "redirectLocation": null, "bodySnippet": null, "slug": "decide" }, "scope": "item", "summary": "Item download looks usable.", "detail": "Yavira can redirect you to the upstream package for this item.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/decide" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] } }, "links": { "detailUrl": "https://openagent3.xyz/skills/decide", "downloadUrl": "https://openagent3.xyz/downloads/decide", "agentUrl": "https://openagent3.xyz/skills/decide/agent", "manifestUrl": "https://openagent3.xyz/skills/decide/agent.json", "briefUrl": "https://openagent3.xyz/skills/decide/agent.md" } } ``` ## Documentation ### Architecture Decision state lives in ~/decide/. If that folder is missing or empty, run setup.md. ~/decide/ ├── memory.md # Durable decision rules, approval boundaries, and confirmed defaults ├── decisions.md # Major decisions with question, components, chosen option, and outcome └── domains/ # Domain-specific decision components, overrides, and exceptions ### When to Use Use when the agent faces a consequential choice that can change architecture, workflow, cost, publish behavior, vendor selection, or long-term project direction. This skill is for branching decisions, not for generic preferences or execution lessons. It should stay compatible with self-improving: self-improving learns how to work better, while decide learns how to choose safely when the choice has lasting consequences. ### Quick Reference TopicFileSetup guidesetup.mdMemory templatememory-template.mdMigration guidemigration.mdDecision componentscomponents.mdConfidence calibrationconfidence.mdExceptions and always-ask casesexceptions.md Use those files as a decision safety stack: first know the structure, then calibrate confidence, then verify exceptions before reusing any past choice. ### Decision Workflow Frame the decision as a real question, not as a vague feeling. Gather the components that materially affect the answer. Read ~/decide/memory.md, then the smallest relevant file in ~/decide/domains/, then check ~/decide/decisions.md for a materially similar record. Reuse a past choice only if the question, the key components, and the exception boundaries still line up. If anything important is missing or changed, ask first and log the answer once the human decides. ### 1. Default Conservative on Consequential Choices Frameworks, architectures, migrations, vendors, publish paths, spending, and irreversible branches should default to asking the human. Safer failure means asking one question too many, not silently picking the wrong branch. A recommendation is good; an unvalidated autonomous choice is not. ### 2. A Decision Requires Components, Not Vibes Every major decision should be framed as a question plus the components that materially affect the answer. Components can include product surface, client type, reversibility, budget, timeline, team size, project constraints, and long-term maintenance cost. If the needed components are missing, the decision is not ready for autonomy. ### 3. Reuse Only When the Context Materially Matches A stored rule is reusable only when the question and the key components match closely enough to make the same choice still rational. Matching one signal is not enough. "Same framework choice" is weak if client, surface, constraints, or risk changed. If there is any serious mismatch, ask first. Exceptions beat defaults. A confirmed default is still invalid when a domain override or high-stakes exception changes the branch. ### 4. Promote Patterns Only After Human Confirmation Repeatedly seeing the same decision in the same context makes a candidate rule, not an autonomous permission. After enough similar decisions, ask: "When this is true plus this plus this, should I default to X?" Only promote the pattern after explicit confirmation. ### 5. Log the Full Decision, Not Just the Outcome Store the question, components, chosen option, rationale, confidence, and outcome. A naked note like "use React Native" is too weak; it must say when and why. Good logs prevent false autonomy later. ### 6. Keep Workspace Routing Non-Destructive Use setup to add small AGENTS and SOUL snippets that force decision retrieval before major choices. Show the exact snippet and wait for explicit approval before writing any workspace file. The routing must make it hard to skip the decision log when a consequential branch appears. ### 7. Never Let Decision Memory Shadow Other Skills Use self-improving for execution quality, corrections, and reusable work habits. Use escalate for ask-vs-act boundaries across actions broadly. Use decide only for major branching choices where the structure of the context determines the answer. ### Common Traps These failures usually come from pattern-matching too early or from collapsing a major decision into a shallow preference. TrapWhy It FailsBetter MoveReusing a rule because the question sounds similarImportant components may have changedCompare question plus key components before reusingTreating one-off emergency choices as defaultsStress decisions rarely generalize wellLog them, but keep them unconfirmed unless repeatedAutodeciding after reading only memory.mdExceptions and domain overrides get missedList domains, read the smallest relevant override, then check decisionsTurning execution preferences into decision rulesBlurs compatibility with self-improvingKeep major branching choices in decide, workflow lessons elsewhereApplying a framework or vendor rule across clients blindlyClient and surface often change the optimal answerAsk again when client, platform, scope, or constraints differ ### Data Storage Local state lives in ~/decide/: durable decision rules, approval boundaries, and confirmed defaults in ~/decide/memory.md major decision records in ~/decide/decisions.md domain-specific component models, overrides, and exceptions in ~/decide/domains/ The packaged guides components.md, confidence.md, and exceptions.md stay in the skill itself and act as references, not as the user's live memory. ### Security & Privacy This skill stores local decision notes in ~/decide/. It may read workspace steering files such as the AGENTS file and SOUL file so that decision retrieval happens before major choices. It may suggest small non-destructive edits to those files during setup, but it must show the snippet and wait for explicit approval before any write. It should prefer asking to guessing whenever a decision can affect money, production, publishing, deletion, contracts, or long-term architecture. It never modifies its own SKILL.md. ### Related Skills Install with clawhub install if user confirms: escalate - Control broad ask-vs-act boundaries around risky actions self-improving - Learn execution lessons without conflating them with decision rules memory - Keep broader long-term context and user continuity proactivity - Push the next step while respecting confirmed decision defaults ### Feedback If useful: clawhub star decide Stay updated: clawhub sync ## Trust - Source: tencent - Verification: Indexed source record - Publisher: ivangdavila - Version: 1.0.1 ## Source health - Status: healthy - Item download looks usable. - Yavira can redirect you to the upstream package for this item. - Health scope: item - Reason: direct_download_ok - Checked at: 2026-05-01T01:10:06.270Z - Expires at: 2026-05-08T01:10:06.270Z - Recommended action: Download for OpenClaw ## Links - [Detail page](https://openagent3.xyz/skills/decide) - [Send to Agent page](https://openagent3.xyz/skills/decide/agent) - [JSON manifest](https://openagent3.xyz/skills/decide/agent.json) - [Markdown brief](https://openagent3.xyz/skills/decide/agent.md) - [Download page](https://openagent3.xyz/downloads/decide)