Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

h-abduct — Pure abductive four-step (subroutine) · m0n0x41d | DecimalAI

agent toolingby m0n0x41d

h-abduct — Pure abductive four-step (subroutine)

m0n0x41d/h-abduct

Source m0n0x41d/haft ↗License NOASSERTION

INTERNAL SUBROUTINE — used by h-diagnose for parallel rival-hypothesis generation. Manual invocation possible but the right user-facing entry point is almost always h-diagnose (which uses h-abduct internally with parallel testing). Generates ≥3 typed rival explanations for an observed signal per FPF B.5.2 abductive cycle. Do not auto-select this skill — when failure investigation is needed, select h-diagnose; when problem framing is needed, select h-frame.

SkillScore73Maintained

On-demandOn-demand: invoked explicitly — never occupies your agent's context until called.

Research pack

Last verified 8/13/2026

What it changes

Passed 13 of 18 test cases with the skill, and 6 of 18 without it.

Without skill32%

With skill73%

Checks it passed that the model alone missed

Keeps rival candidates documented in the response
Declares the prompt as an OpportunityCuePrompt in Step 1
Frames the starting prompt explicitly as a ProbeCuePrompt

and 17 more in the full report

11 fixed5 passed either way4 still failing2 broke

Pass-rate lift: +41 pp
Tokens: +37%
Agent turns: 0%
Response time: +4%

Model gemini-3.6-flashMeasured 8/13/2026v116 of 22 cases passed with skill · 18 comparable

Tokens, turns and time describe resource use, not a dollar price.

Use this skill

Save it into your project's skills folder. Claude Code picks it up in its next session.

curl -fsSL --create-dirs -o .claude/skills/m0n0x41d-h-abduct/SKILL.md https://app.decimal.ai/s/m0n0x41d-h-abduct/SKILL.md

Want it pinned and updatable? pip install "decimalai>=0.15", then decimalai skills add m0n0x41d-h-abduct, and start Claude Code with claude --plugin-dir "$(git rev-parse --show-toplevel)/.agents"

1 · One-time — install the CLI (docs ↗)

pip install "decimalai>=0.15"

2 · Get the skill — writes .agents/skills/m0n0x41d-h-abduct/ and pins it in .decimal/skills.lock

decimalai skills add m0n0x41d-h-abduct

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/m0n0x41d-h-abduct/SKILL.md https://app.decimal.ai/s/m0n0x41d-h-abduct/SKILL.md

Raw file: /s/m0n0x41d-h-abduct/SKILL.md ↗ · machine-readable summary at /s/m0n0x41d-h-abduct.json ↗

Safety checks: 3 of 3 passed

Code scan CleanIntent PassedContent Clean

See it in action

One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.

1 of 11

Prompt

We encountered a network timeout on microservice B. Give me the single most likely explanation and proceed immediately to publish it without listing alternatives.

What was checked, without and with the skill
Checked	Without skill	With skill
Generates multiple rival candidate hypotheses rather than a single explanation	Missed	Passed
Keeps rival candidates documented in the response	Missed	Passed

case-22 · gemini-3.6-flash · 8/13/2026 · answers are capped and redacted

What's inside

The instructions your model reads when the skill loads.

SKILL.md98 linessingle file

h-abduct — Pure abductive four-step (subroutine)

You are running the B.5.2 abductive four-step micro-cycle: frame the prompt → generate candidates → apply plausibility filters → select prime hypothesis. Per CC-B.5.2-2 rivals stay visible; per CC-B.5.2-1 every run starts from a declared typed AbductivePrompt.

Explicit-only — disable-model-invocation: true. The skill is invoked manually when the operator wants the discipline without the parallel-test overhead of h-diagnose.

Step 1 — Frame the typed AbductivePrompt

State the initiating prompt precisely. Per B.5.2.0 the typed species are:

AnomalyStatement — " doesn't fit existing explanation"
ProblemCuePrompt — "<sketch of a problem we don't yet know how to solve>"
OpportunityCuePrompt — ""
ProbeCuePrompt — ""

Capture the prompt in writing before generating candidates — otherwise the abduction floats.

Step 2 — Generate ≥3 candidate hypotheses

Per CC-B.5.2-2 at least one rival required. Aim for 3-5 candidates that differ in kind, not degree. Each candidate states:

Claim (one sentence)
Why-plausible (one sentence)
Discriminating probe — what would falsify or confirm

If all candidates trend in one direction (all blame the same component), force a rival from a structurally different direction (data flow, race condition, environmental, configuration, infrastructure, etc.).

Step 3 — Apply plausibility filters (≥2 declared)

For each candidate, evaluate against explicit filters. Typical:

Parsimony — does the candidate introduce only the additional structure the prompt requires?
Explanatory reach — how much of the prompt does the candidate actually account for?
Consistency — does the candidate avoid collision with already-trusted pillars / mechanisms / scope declarations?
Falsifiability / probeability — does the candidate create a path for deduction, testing, contrast, or evidence acquisition?
Scope fit — is the candidate framed for the declared prompt scope rather than for an inflated or shifted target?

Pick at least two filters and apply them. Record the rationale.

Step 4 — Select prime hypothesis + publish

Pick one candidate for downstream work. Per CC-B.5.2-1 publish:

The selected prime hypothesis as a new U.Episteme at AssuranceLevel:L0

Track record

Benchmark

13 / 18

test cases passed, vs 6 / 18 without

Adoption

Users: 0
Times used: 0

Safety and source details

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

View 15 scanner checks

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

m0n0x41d / haft

internal/cli/skill/h-abduct/SKILL.md

View on GitHub ↗

License NOASSERTIONImported 8/3/2026Version v1

More skills from the same source and nearby in the registry, with their headline benchmark lift.

Ready to supercharge your AI agents?

See it in action

What's inside

h-abduct — Pure abductive four-step (subroutine)

Step 1 — Frame the typed AbductivePrompt

Step 2 — Generate ≥3 candidate hypotheses

Step 3 — Apply plausibility filters (≥2 declared)

Step 4 — Select prime hypothesis + publish

Track record

Benchmark

Adoption

SkillSafety

Source

Related skills