Ready to supercharge your AI agents?

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

Parallel Analysis and Synthesis · trpc-group | DecimalAI

agent toolingby trpc-group

Parallel Analysis and Synthesis

trpc-group/fanout-analysis

Source trpc-group/trpc-agent-go ↗License Apache-2.0

Use this workflow recipe when a question benefits from several independent perspectives followed by a single evidence-based synthesis.

SkillScore91Maintained

Automatic

AI Engineer pack

Last verified 9/29/2026

What it changes

Passed 20 of 22 test cases with the skill, and 10 of 22 without it.

Without skill45%

With skill91%

Checks it passed that the model alone missed

The code explicitly tests for None branch returns.
The synthesizer output schema includes a reasons field.
The synthesizer receives explicit notice of any missing branch outputs.

and 18 more in the full report

11 fixed9 passed either way1 still failing1 broke

Pass-rate lift: +45 pp
Tokens: +7%
Agent turns: 0%
Response time: -14%

Model gemini-3.6-flashMeasured 9/29/2026v120 of 22 cases passed with skill

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/trpc-group-fanout-analysis/SKILL.md https://app.decimal.ai/s/trpc-group-fanout-analysis/SKILL.md

Want it pinned and updatable? pip install "decimalai>=0.15", then decimalai skills add trpc-group-fanout-analysis, 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/trpc-group-fanout-analysis/ and pins it in .decimal/skills.lock

decimalai skills add trpc-group-fanout-analysis

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

curl -fsSL --create-dirs -o .claude/skills/trpc-group-fanout-analysis/SKILL.md https://app.decimal.ai/s/trpc-group-fanout-analysis/SKILL.md

Raw file: /s/trpc-group-fanout-analysis/SKILL.md ↗ · machine-readable summary at /s/trpc-group-fanout-analysis.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 are deciding whether to refactor our legacy monolith into a serverless architecture or stick to a modular monolith model. Please evaluate this decision concurrently from three distinct perspectives: infrastructure cost, developer velocity, and operational reliability. Synthesize these viewpoints into a clear decision output that gives us a final recommendation, the primary supporting reasons, and any key uncertainties that need further validation.

What was checked, without and with the skill
Checked	Without skill	With skill
The implementation uses parallel execution constructs to run the analysis branches concurrently.	Missed	Passed
The workflow creates dedicated Agent instances for each individual analysis branch.	Missed	Passed
The synthesizer extracts a structured output object with a recommendation field.	Missed	Passed

case-01 · gemini-3.6-flash · 9/29/2026 · answers are capped and redacted

What's inside

The instructions your model reads when the skill loads.

SKILL.md50 linessingle file

Parallel Analysis and Synthesis

Turn the user's request into a temporary fan-out/fan-in workflow. Keep the number and focus of branches appropriate to the request; do not create roles just to make the workflow look larger.

Process

Extract the decision or question, the relevant constraints, and the output format from the user's request.
Choose two to four independent analysis angles that cover different evidence or reasoning needs. Give every branch the same core question and only the context it needs. Do not let one branch depend on another branch's unfinished answer.
Run those branches in parallel. Each branch should return concise findings, assumptions, and unresolved uncertainty rather than a polished final answer.
Pass the ordered branch results, the original request, and the explicit decision criteria to a separate synthesis role. The synthesizer must distinguish agreement, disagreement, and missing evidence; it must not silently turn an unsupported claim into a fact.
If the request requires a decision, have the synthesizer return a small structured object with recommendation, reasons, and uncertainties. Keep branch content as text unless a later control-flow decision genuinely needs typed fields.
Return the synthesis and the key evidence trail. If a branch fails, keep that missing evidence explicit and follow the application's bounded failure policy; do not silently treat it as support or retry indefinitely.

Compilation Rules

Express independent branches with parallel([...]); preserve the input order when passing results to the synthesizer.
Use separate workflow-local Agent instances for each branch and for the synthesizer. Do not ask the synthesizer to redo every branch from memory.
Pass the original question, constraints, and branch outputs explicitly as inputs. A later stage must not depend on context that was only present in a previous Agent's prompt.
parallel returns None for a failed independent branch. Handle that value explicitly, and let the workflow or its caller decide whether a single,

Track record

Benchmark

20 / 22

test cases passed, vs 10 / 22 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 v5 — 8/31/2026

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

trpc-group / trpc-agent-go

examples/dynamicworkflow/skills/skills/fanout-analysis/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 8/31/2026Version 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

Parallel Analysis and Synthesis

Process

Compilation Rules

Track record

Benchmark

Adoption

SkillSafety

Source

Related skills