Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/zimalabs/code-decisions/decisionnpx skills add zimalabs/code-decisions --skill decisiongit clone --depth 1 https://github.com/zimalabs/code-decisionsWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00023 | $0.00904 |
| Opus 5 | $0.00012 | $0.00452 |
| Sonnet 5 | $0.00005 | $0.00181 |
| Haiku 4.5 | $0.00002 | $0.00090 |
Grade A, and why
decision scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/decision
Parse $ARGUMENTS to determine intent and route to the appropriate action.
CLI Usage
The SessionStart hook injects a Decision CLI: ... line into conversation context with the correct PYTHONPATH prefix. Always use that exact prefix when running CLI commands. For brevity, this document writes python3 -m decision <command> — prepend the injected PYTHONPATH="..." when executing.
Intent Detection
Explicit subcommands take priority — /decision search <query> and /decision capture <statement> bypass intent detection entirely.
Search (default): questions, keywords, lookups, --tags, --stats, --coverage, or anything ambiguous.
Capture: declarative statement with both a choice and a reason ("chose X because Y", "going with X instead of Y"). Without a reason, prefer search — the user may be looking up an existing decision.
Manage: admin verbs — edit, undo, dismiss, debug, review, enrich, tree, tour, history, help.
No arguments: show brief status with python3 -m decision stats --json and suggest quick actions.
Search
- Check pre-seeded results first — a hook runs FTS5 search automatically when
/decision <keywords>is invoked. If results exist in conversation context, present them directly. - If pre-seeded results are empty or insufficient: use
Grepto search.claude/decisions/files for keywords, thenReadmatching files to present results. This avoids permission prompts. Only fall back topython3 -m decision search <keywords>via CLI if native tools fail. - Flags:
--tags→python3 -m decision tags,--stats→python3 -m decision stats,--coverage→python3 -m decision coverage
Capture
- Search existing decisions first — if one covers the same topic, edit it in place instead of creating a new file.
- Write directly to
.claude/decisions/{slug}.mdfollowing the decision template (injected at session start). The content-validation hook will reject and guide you if anything is wrong. - Confirm briefly:
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 70 lines · 23 tokens per session scan A 68855066eb21
decision is a skill published in the GitHub repository zimalabs/code-decisions (5 stars, last pushed 1mo ago), licensed MIT. It adds 23 tokens to every session and 904 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
adr-manager
Record and retrieve Architecture Decision Records — log a decision's context, options, rationale, and tradeoffs, then list/search them. Use when making a non-trivial design or architecture decision, choosing between tools/approaches, or asked 'why did we do it this way?'.
adr-writing
Write Architecture Decision Records future engineers will actually read. Invoke when making a non-trivial architectural choice that you want to defend later.
adr
Architecture Decision Record (ADR) — Philosophy-First: Creates well-structured ADRs using a 3-phase process: (1) Define the architectural philosophy and identity of the system, (2) Explore options through the lens of that philosophy, (3) Adversarial review by a devil's advocate agent. Use this skill whenever the user…
phase-5-adr
Phase 5 — Architecture Decisions with ADR generation and validation.
openlore-brainstorm
Transform a feature idea into an annotated story using a Domain Sketch or Constrained Option Tree. Use when asked to brainstorm, explore, or shape a feature before implementation.
openlore-execute-refactor
Apply a confirmed .openlore/refactor-plan.md with a test gate after each change. Use when asked to execute or continue an OpenLore refactoring plan.