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.
git clone --depth 1 https://github.com/davesheffer/hunchWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/rules/davesheffer/hunch/hunch)<a href="https://agentmods.dev/rules/davesheffer/hunch/hunch"><img src="https://agentmods.dev/badge/rules/davesheffer/hunch/hunch/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/rules/davesheffer/hunch/hunch"><img src="https://agentmods.dev/badge/rules/davesheffer/hunch/hunch.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00989 | $0.00989 |
| Opus 5.5 | $0.00396 | $0.00396 |
| Sonnet 5.5 | $0.00198 | $0.00198 |
| Haiku 4.5 | $0.00099 | $0.00099 |
Grade A, and why
hunch 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 5d ago.
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 — 32 lines — stays where its author put it; the contents beside it link to each section on GitHub.
🧠 Hunch (Engineering Memory)
This repo has Hunch, a graph of why the code is the way it is. It holds 339 decisions, 2 bugs, 31 constraints, 22 components, 3 policies, 26 open findings. Use the hunch_* MCP tools by moment:
- Start: reuse the task ID and
task verifycommand the prompt hook printed; with none, callhunch_task(action: "start", title)once. Thenhunch_context(target, task_id)first. Orient withhunch_structure,hunch_workspaces,hunch_runbook(task). Ask the user about eachhunch_escalations()entry; silence is never approval. - Design:
hunch_why(target)(includes what was rejected),hunch_current_decision(topic),hunch_bug_lineage(symptom_or_symbol),hunch_compare(candidates),hunch_query(query). - Edit:
hunch_check_constraints(scope),hunch_get_dependents(symbol)/hunch_blast_radius(target),hunch_findings(scope?). - Merge:
hunch_conformance(),hunch_pr_impact(base?),hunch_merge_verdict. Policy review:hunch_policy_evaluate,hunch_policy_plan(policy_id),hunch_policy_card(policy_id),hunch_policy_proof; only a human activates a policy. - Record:
hunch_capture_decision→hunch_record_decision;hunch_record_correctionturns a human correction into an enforced rule;hunch_record_findingkeeps an observation with evidence. Pass the task_id. - Finish: run checks through the
task verifylauncher. If the task used Hunch, you started it, or no host stop hook closes it, callhunch_task(action: "finish", task_id)and show its card verbatim. Itsapplicationsschema carries the claim rules. - To update Hunch, run
hunch updatefrom the repo root.
⛔ Top invariants (do not break)
- [blocking] Synthesis must use an explicitly selected coding-assistant subscription CLI or the deterministic local fallback; never call a pay-per-token API (scope: src/synthesis/**; con_2ce3f2a547)
- [blocking] This repo's generated wiki must never be committed to the public repo — private overlay wiki only (hunch wiki --private); wiki/ stays gitignored (scope: wiki/**; con_547fff76bd)
- [blocking] Config/provider writers must merge idempotently into existing user files and refuse to clobber an unparseable file (scope: src/integrations/claudeConfig.ts, src/integrations/claudemd.ts, src/integrations/providers.ts, src/integrations/scaffold.ts; con_8460b6770f)
- [blocking] All JSON writes to .hunch/ must be atomic (temp-file + rename) so an interrupted write can never truncate the index (scope: src/store/**, src/core/io.ts; con_902759b3dc)
- [blocking] Forward-migrate raw JSON to the current schema BEFORE Zod validation; never silently drop unmigratable records (scope: src/core/migrate.ts, src/store/jsonStore.ts; con_947c578b2c)
- [blocking] Semantic vectors are a derived layer in SQLite, never the source of truth; reconcile by content hash on reindex (scope: src/store/**; con_a87360128b)
- [warning] Replace the blue-violet palette in the homepage particle helicoid with the prior green Hunch palette; convey gene mapping through structured loci, paired bands, quiet gaps, and moving mapped regions… (scope: site/dna-hero.js; con_009f720549)
- [warning] Memory hygiene is the agent's job, done without waiting for the human: accept shipped roadmap proposals, settle provenance-repair escalations (apply the true commit or drop), re-anchor or stale… (scope: .hunch/**, ROADMAP.md, docs/autonomous-development.md; con_039cee7367)
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.
- 5d ago Changed 0f0ac249549e
- 16d ago Changed bef87d549586
- 19d ago Changed f8a1ffc73ff6
- 20d ago Changed 54d4a8f8d48c
- 21d ago Changed · -36 lines · -1,393 tokens per session bfb82935ded8
- 24d ago Changed 14a8cd30560a
- 26d ago Changed · +105 tokens per session de79266acacb
- 27d ago Changed cb2fb7b24a59
- 29d ago Changed · +1 lines · +61 tokens per session f770f9d658ee
- 1mo ago First seen · 67 lines · 2,216 tokens per session scan A fb404f6954ba
hunch is a cursor rule published in the GitHub repository davesheffer/hunch (11 stars, last pushed 5d ago), licensed Apache-2.0. It adds 989 tokens to every session, about $0.0040 per session on Opus 5.5. 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 cursor rules, from other repositories
dreamd-recall
Recall lessons, decisions, and prior context from the .agent/ memory daemon. Use when starting work in a project that has a .agent/ folder, when the user references a past decision, or when you are about to make a choice that has a documented prior.
ai-dev-os-review
Performs a comprehensive self-review before creating a PR. Combines guideline compliance checking (L3) with design-level review (L2) and philosophical alignment (L1). Unlike @ai-dev-os-check which only checks rules, this also evaluates architecture decisions and code design quality.
context-delegate
Context Gatherer & External LLM Prompt Generator.
claude-skills-karpathy
Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.
ponytail
Ponytail, lazy senior dev mode. Always pick the simplest solution that works.
angular-20
This rule provides comprehensive best practices and coding standards for Angular development, focusing on modern TypeScript, standalone components, signals, and performance optimizations.