orchestrate-banking-brief-pipeline

Orchestrate a four-stage banking BA brief pipeline with validated JSON handoffs and pluggable stage commands.

Updated Apr 26, 2026
One-click install
npx skills add https://github.com/Jin9/skillify-foundation --skill orchestrate-banking-brief-pipeline-jin9
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: orchestrate-banking-brief-pipeline
Source: https://github.com/Jin9/skillify-foundation/tree/main/treasury/orchestrate-banking-brief-pipeline
Command: npx skills add https://github.com/Jin9/skillify-foundation --skill orchestrate-banking-brief-pipeline-jin9

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires jsonschema, and includes scripts (resource) and references (resource) components.

What problem does it solve? Coordinating multiple business-analysis skills (extraction, compliance, ambiguity sweep, Gherkin generation) by hand produces inconsistent handoffs, lost traceability, and silent failures. This Skill chains the four decomposed banking brief stages into one deterministic pipeline with schema-validated JSON handoffs, centralized state, and explicit failure reporting. ## Core Features & Use Cases - Ordered Multi-Stage Execution: Runs extraction, compliance, ambiguity, and Gherkin stages in a fixed order, feeding each stage the original input plus all prior validated outputs. - Deterministic Validation & State: Validates pipeline input and every stage output against JSON schemas, appends results to pipeline_state.json after each attempt, and halts on command failure, invalid JSON, schema errors, or blocked stage status. - Provider-Agnostic Runner: Executes operator-supplied stage commands with {input}, {output}, {skill_dir}, {stage}, and {state} placeholders, without calling any model API directly. - Use Case: A business analyst receives raw Jira content about wire transfer status notifications and needs a compliance-checked brief with Gherkin acceptance criteria. The pipeline validates the input, runs all four stages, and writes output.json marked as brief or blocked_partial_brief depending on compliance blockers and P1 open questions. ## Quick Start Run the decomposed banking brief pipeline on my input JSON file using the four configured stage commands and write the assembled output to my chosen output directory.

Frequently Asked Questions about orchestrate-banking-brief-pipeline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run the banking brief pipeline end to end?▼

Run scripts/run_elicitation_pipeline.py with --input pointing to a valid pipeline input JSON, --output-dir for results, and one --stage-command per stage in key=command form for extract, compliance, ambiguity, and gherkin. The runner validates input, executes stages in order, and writes output.json on success.

What input format does the banking brief pipeline require?▼

The input must be JSON matching templates/pipeline_input.schema.json with required fields raw_content (200-200000 characters) and a UUID idempotency_key. Optional fields include source_type, tier_hint, requested_at, and project_context.

Does the pipeline call an LLM or vendor API directly?▼

No, the runner is provider-agnostic and never calls a model API itself. Each stage is an operator-supplied shell command with placeholders for input, output, skill directory, stage name, and state path, so you plug in whatever model invocation you trust.

What happens when a pipeline stage fails or returns blocked status?▼

The orchestrator halts immediately and writes failure.json containing failed_stage, failure_code, failure_reason, and evidence such as the command attempt or validation errors. It also records the failure in pipeline_state.json so a human can diagnose and retry.

When is the pipeline output marked as blocked_partial_brief?▼

The assembled output.json sets output_type to blocked_partial_brief and blocks_tl_handoff to true when the compliance stage reports blockers or any P1-severity open question remains from the ambiguity sweep. Otherwise the output_type is brief.

Is it safe to configure arbitrary stage commands in the pipeline?▼

Stage commands run through the shell with the operator's full privileges and are not sandboxed or escaped by the runner. Configure only commands you trust and quote paths containing shell metacharacters, as documented in the orchestration contract.