siftcoder
/siftcoder
Wskazuje, który skill SiftCoder pasuje do danego zadania, na podstawie drzewa decyzyjnego z kilku rodzin.
- Autor
- ialameh ↗
- Repozytorium
- sift-coder ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 03.06.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 03.06.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Wskazuje, który skill SiftCoder pasuje do danego zadania, na podstawie drzewa decyzyjnego z kilku rodzin.
Wyzwalacze z opisu
Oryginalny opis (EN)
Use as a meta-skill — "which SiftCoder skill should I use for X". Decision tree across families. Use when user is unsure which specific skill applies.
Jak działa
- Quick triage
- Method
- Rules
- Anti-patterns
- Model-drift audit (review every 3–6 months)
- Value over native CC
Podgląd SKILL.md
---
name: siftcoder
description: Use as a meta-skill — "which SiftCoder skill should I use for X". Decision tree across families. Use when user is unsure which specific skill applies.
---
# siftcoder (meta)
Decision tree across SiftCoder skill families. Use when the user has a goal but isn't sure which skill fits.
## Quick triage
| User says | Skill |
|---|---|
| "build me X" | coding/build |
| "fix this bug" | coding/fix |
| "test this" | coding/tdd |
| "investigate why" | coding/investigate or reasoning/archaeologist |
| "refactor" | coding/refactor |
| "make it faster" | coding/optimize |
| "make it simpler" | coding/zen |
| "review this PR" | review/review (with built-in /review) |
| "is it secure" | review/security or built-in /security-review |
| "is it compliant" | quality/comply or salesforce/salesforce-comply |
| "what does this do" | reasoning/narrator |
| "why does this exist" | reasoning/archaeologist |
| "find the pattern" | knowledge/pattern-search |
| "save this pattern" | knowledge/pattern-learn |
| "what could break" | quality/blast-radius or quality/chaos |
| "find edge cases" | quality/fuzz-mind |
| "show me what changed" | docs/codemap-diff |
| "explain how this works for a non-eng" | reasoning/narrator |
| "ideas for next features" | ux/ideate |
| "new project ideas" | ux/surprise-me |
| "spec this out" | spec/improve-spec or spec/spec-from-stories |
| "what's missing" | spec/gap-analysis |
| "is this doable" | spec/feasibility |
| "memory capture this" | knowledge/pattern-learn or `mem_drain` |
| "memory query" | `mem_search` (MCP tool, no skill needed) |
| "Salesforce X" | salesforce/* |
## Method
1. **Listen.** What did the user actually say?
2. **Match.** Use the table above as a starting point.
3. **Ambiguous?** Ask: "Sounds like you want X — should I use the `<skill>` skill, or is it more like Y?"
4. **Hand off.** Don't solve in this skill — invoke the matched skill.
## Rules
- **Don't solve in meta.** Always hand off to the specific skill.
- **One skill at a time.** Don't try to apply 3.
- **When ambiguous, ask once.** Then commit.
- **Memory the matches.** Future similar phrasings auto-route.
## Anti-patterns
- Doing the work in the meta skill
- Inventing a skill that doesn't exist
- Hedging ("could be A or B")
## Model-drift audit (review every 3–6 months)
Instructions and pins tuned for today's model can work against a future one. On every major model release, re-check the model-coupled surfaces and retire anything compensating for an obsolete limitation:
| Surface | What to check |
|---|---|
| `settings.json` → `summarizer.modelHaiku` / `modelSonnet` | Still the right tier? Still a current, non-retired model ID? |
| `settings.json` → `summarizer.confidenceThreshold` (0.6) | Recalibrate against current Haiku quality — a smarter Haiku may warrant a higher bar before Sonnet escalation. |
| `ollama.summarizeModel` / `embedModel` | Newer local models may beat the pinned ones on cost/quality. |
| Skill/agent text that hard-codes model behavior | Wording written around an old limitation ("the model can't…") may now be wrong — delete it. |
The pins live in `settings.json` (`siftcoder.memory.summarizer` / `siftcoder.ollama`); `siftcoder mem doctor` reports backend reachability. This is hygiene, not a feature: schedule it, don't wait for a regression.
## Value over native CC
CC chooses paths organically. This skill is the **explicit decision tree** — useful when the user wants the "official" SiftCoder skill for their phrasing. Speeds discovery.
Wersje
W bazie nie ma innych skilli o nazwie siftcoder.
Migawka bazy z 13.11.2026 · wersja paczki 955098b79022 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100prompt
Pomaga napisać lepszy prompt dla skilli SiftCoder, modelu językowego lub człowieka, z uwzględnieniem kontekstu SiftCoder. - 100memory-usage
Odpytuje trwałą pamięć projektu SiftCoder przez narzędzia mem_search, mem_get, mem_timeline, mem_why i mem_drain, aby ustalić, co i dlaczego zdecydowano. - 100using-agent-skills
Wyszukuje i uruchamia skille agentów, wybierając, który skill lub przepływ pasuje do zadania na początku sesji. - 100using-agent-skills
Wykrywa i wywołuje skille agentów na początku sesji oraz wtedy, gdy trzeba ustalić, który skill pasuje do bieżącego zadania. - 100using-agent-skills
Wyszukuje i uruchamia skille agentów, wybierając, który skill lub przepływ pasuje do zadania na początku sesji. - 100using-agent-skills
Wyszukuje i uruchamia skille agentów, wybierając, który skill lub przepływ pasuje do zadania na początku sesji.