SinFire splits clarifying a task into three steps: align terms, research facts in parallel, then settle decisions in dependency order, so flagship models stop grilling you on trivia
The popular grill-me skill (Matt Pocock) makes an agent interview you before it plans; on today's flagship reasoning models the original 'relentlessly' often means a dozen nitpicky questions. Chinese-community author SinFire (@Sin_Yiyao) reworked grill-me plus his own requirements-clarification practice into sinfire-skills (MIT, v0.1.0): three instruction-only skills, 119 lines of Markdown, no runtime deps. $semantic-clarification (59 lines) only asks what you mean, and will not expand into requirements gathering. $facts-searching (26 lines) dispatches parallel sub-agents to search the web, dedupes, resolves conflicts, and brings back only sourced findings that affect the current decision—unverifiable items stay marked as assumptions. $clarify-decisions (34 lines) stays at the current level, defaults low-impact reversible details, asks only the most consequential frontier questions each round, and recommends an answer only when evidence and stated aims support one. Install: npx skills add SinVC/sinfire-skills --skill clarify-decisions semantic-clarification facts-searching -a codex. We cloned the repo and bun run validate passed for all three skills plus the Codex explicit-invocation policy; we did not run a live Codex/Claude session through the full loop.
Visit site sinfire-skills ↗- Pricing
- Free
- Region
- Global
- Where it's from
- Chinese community
- Added
- Nov 8, 2026
- Tested on
- Nov 8, 2026
- Version we used
- sinfire-skills 0.1.0 @ d445163; Bun 1.4.2 (bun run validate)
- Categories
- CodingProductivityResearch
- Original author
- SinFire @Sin_Yiyao ↗
- Our test
- Partly tested
- Cost / time
- Repo is MIT/free. Needs a SKILL.md-capable agent (author example is Codex; the Skills CLI also supports Claude Code and others). facts-searching needs web search and parallel sub-agents; when the host allows, it prefers a faster, cheaper worker model. No extra API.
The task
Before anyone writes an implementation plan, align what the words mean, look up the facts that can be looked up, and settle high-impact irreversible decisions in dependency order—without an agent grilling you on trivia.
Steps
- In the project where you want the skills: npx skills add SinVC/sinfire-skills --skill clarify-decisions semantic-clarification facts-searching -a codex. Install all three for the full loop; installing one does not pull in companions. These are instruction-only—no Bun, no npm publish. The Skills CLI supports other agents too.
- Recommended order: $semantic-clarification first for terms, ambiguity and unclear references that could change the task; $facts-searching when you need outside evidence; then $clarify-decisions to settle consequential decisions in dependency order. clarify-decisions is explicit-invocation only.
- Paste: `$clarify-decisions Help me clarify the decisions behind this product idea before planning implementation.` For terms: `$semantic-clarification Identify terms or references in this brief that could materially change its interpretation.` For research: `$facts-searching Research the facts needed to compare these options, then report conclusions with sources and remaining uncertainty.`
- Stay at the current level: do not drop from requirements into implementation unless asked. Default low-impact reversible details. When the choice turns on values, priorities or risk tolerance, lay out the trade-off and let the user choose. Anything unverifiable stays an assumption, never a silent basis for an irreversible call.
- You should end with a short record: what was decided, what was verified (and where), and what is still uncertain. The author says the agent should not act on the result until you accept it.
Prompt
$clarify-decisions Help me clarify the decisions behind this product idea before planning implementation.
What the author says
Author @Sin_Yiyao: grill-me's 'relentlessly' makes flagship reasoning models ask a dozen nitpicky questions; sinfire-skills drops indiscriminate grilling, stops forcing a recommended answer on every question, splits fact-finding into parallel sub-agents, and aligns terminology first. Three skills, 119 lines, no runtime deps. He marked the 'better than grill-me' claim with 'Maybe' because v0.1.0 launched with no stars yet, but said it has been more usable than the original in his own work.
Our test
Cloned SinVC/sinfire-skills (commit d445163, package.json version 0.1.0). Read the bilingual README, all three SKILL.md files, and clarify-decisions/agents/openai.yaml. Ran bun run validate with local Bun 1.4.2. Did not install into Codex/Claude Code or run a live clarification session.
What we found when we tried it:validate printed that it verified 3 skills: clarify-decisions, facts-searching, semantic-clarification. SKILL.md line counts were 34, 26 and 59 (119 total), matching the author's claim. openai.yaml sets allow_implicit_invocation to false. CHANGELOG still labels 0.1.0 as Unreleased.
How to get in
Public GitHub, MIT, with English and Chinese READMEs. The author's install example targets Codex (-a codex); the Skills CLI can install into other compatible agents. The original write-up is in Chinese.
Who it suits
People who like grill-me but find flagship models over-question, and who want meaning and facts separated before a plan. Skip if you don't run Codex/Claude Code or don't want an extra clarification round.
Pitfalls
- clarify-decisions is explicit-invocation only and will not jump in unasked; installing one skill does not guarantee companion skills are present.
- facts-searching needs web search and parallel sub-agents; without them, clarify-decisions investigates itself, with different quality and cost.
- This is v0.1.0—CHANGELOG still says Unreleased—and the author hedged 'better' with 'Maybe'. Judge it on your own session.
- Our run only validated the release layout; we did not walk a live agent through align → research → decide.
More in this category
- A demo video per docs page that flags itself when the UI changesVideoScreenplayClaude CodeTool free; uses your agent's quota◐ Partly testedOverseas
- AI backend code that guards against races, leaks and bad migrationsCodingbackendskillsClaude CodePostgreSQLFree; about 2× tokens per task◐ Partly testedOverseas
- An afternoon of AI work, a 3D action game you play in the browserCodingClaude CodeTripoJimengSkill free; Claude quota plus asset tools◐ Partly testedChina
- Keep a long-running AI-built project from turning into a messCodingClaudeCodexoil-delegateNeeds Claude and Codex plans◐ Partly testedChina
- Use a web app once and get an AI write-up of how it worksResearchREAClaude CodeChromeFree tools; needs Claude or Codex◐ Partly testedOverseas