A framework for building large-scale, deterministic, interactive workflows with a fault-tolerant, conversational UX
Skill Claude CodeCodex
Load this skill when you need to understand or preserve fastWorkflow's load-bearing runtime design before changing core code: anything touching WorkflowExecutionContext, ChatSession, the CME/wildcard NLU pipeline, askuser suspension/resume, AskUserSuspend / CommandCancelledError, SessionStateStore…
Skill Claude CodeCodex
Load this skill when setting up the fastWorkflow dev environment from scratch or fixing a broken one: fresh clone, "which Python / poetry install / pip extras do I need", "tests are all skipping", "FileNotFoundError threshold.json", "make gen-env fails", "where do API keys go", "env/.env vs passwords/.env vs…
Skill Claude CodeCodex
Load this skill BEFORE making any change to the fastWorkflow repo that touches git, beads (bd), version numbers, tests, or documents in docs/. Triggers: "commit this", "push", "bump the version", "close the issue", "bd create/update/close", "delete this test", "clean up docs/", "should this get a design doc?", "is…
Skill Claude CodeCodex
Load this skill whenever you touch fastWorkflow configuration: any LLM / LITELLMAPIKEY / SPEEDDICTFOLDERNAME / INTENTDETECTION / SESSIONSTATE env var, the fastworkflow.env or fastworkflow.passwords.env files, a fastworkflow CLI subcommand or flag, LiteLLM proxy routing, or JWT server settings. Trigger symptoms: "env…
Skill Claude CodeCodex
Symptom-to-triage playbook for fastWorkflow runtime, build, and train failures. Load this when you see: the wrong command being matched, "threshold.json not found" / FileNotFoundError in commandinfo, the internal CME workflow retraining on every fastworkflow train, parameter extraction looping on NOTFOUND, the CLI…
Skill Claude CodeCodex
Load this skill when you need to MEASURE a fastWorkflow workflow instead of eyeballing it — inspect commandinfo trained artifacts, check whether cached command snapshots are stale (fingerprint FRESH/STALE), capture or read a turn's command traces / turn records from observability.sqlite3, smoke-test intent-classifier…
Skill Claude CodeCodex
Load when writing, editing, or citing any fastWorkflow documentation or public claim — README changes, design docs, articles, benchmark numbers, DOIs, "which doc is authoritative?", "can we say/publish X?", "the README/CLAUDE.md says...", starting a new design doc or design review, or handling the team-private docs in…
Skill Claude CodeCodex
Load this when you are about to investigate, revert, re-attempt, or "clean up" something in fastWorkflow and suspect a battle may already have been fought: symptoms like "this bug looks old", "why does this dead code / legacy file exist", "should we just delete X", "the revert didn't revert", "bd close didn't stick"…
Skill Claude CodeCodex
Load when tuning a workflow's intent-detection quality or judging whether one training change beat another — trigger phrases include "intent detection is weak", "should I add more utterances", "how many seed utterances per command", "did more seeds help", "training F1 is high but routing is bad", "what is…
Skill Claude CodeCodex
Domain reference for fastWorkflow's NLU stack AS IMPLEMENTED IN THIS REPO: two-tier TinyBERT/DistilBERT intent classifiers, confidence thresholds, synthetic utterance generation (litellm + PersonaHub), DSPy parameter extraction, runtime matching layers (fuzzy/embedding-cache/classifier), and per-role litellm model…
Skill Claude CodeCodex
Load when you are about to propose, run, review, accept, or retire a research idea or experiment in this repo — trigger phrases include "I have a hypothesis", "let's try adding X and see if the score improves", "the numbers went up so it works", "should we pin the simulator temperature", "write up this experiment"…
Skill Claude CodeCodex
Operate fastWorkflow at runtime: use when you need to run the CLI (examples/build/refine/ train/run/runfastapimcp), fetch or pick a bundled example workflow, do the quickstart, understand agent vs assistant vs "/" deterministic mode, identify or clean runtime artifacts (commandinfo, workflowcontexts, convoinfo…
Skill Claude CodeCodex
What counts as evidence in fastWorkflow: the real state of the 1803-test pytest suite, which tests skip/no-op and why, the golden test workflows and trained artifacts you must never destroy, how to add integration tests, and the acceptance bar before a release tag. Load this when you are about to run the test suite…
Skill Claude CodeCodexCursor
Uses bd (beads) for issue tracking. Use when finding work, claiming or closing issues, syncing with git, or when the user mentions bd, beads, issues, or tasks.
Skill Claude CodeCodexCursor
Completes work sessions by running quality gates, syncing bd, pushing to remote, and handing off. Use when ending a work session, landing the plane, session completion, or when the user asks to wrap up or finish work.
Skill Claude CodeCodex
Build a two-tier conversation benchmark for a fastWorkflow workflow: short single-errand conversations that act as unit tests and reusable blocks, then long real-world tasks of 20-100 steps composed from those blocks. Covers turn and conversation structure, seeding objects with enough structure to sustain a chain…
Skill Claude CodeCodex
Diagnose fastWorkflow task failures from the conversation logs in a workflow's observability.sqlite3: locate and safely read the database, walk a turn's span trace to the failing pipeline stage (intent routing, parameter extraction, dblookup, validation, planning, askuser, or the app's own code), and route each…
Skill Claude CodeCodex
Declare availablefrom producer hints on fastWorkflow command parameters so the planner can chain commands into a working sequence and so a missing parameter tells the caller which command supplies it. Covers the hint format, how the planner and the missing-parameter message consume it, choosing a producer the caller…
Skill Claude CodeCodex
Design and tune the context inheritance (base) and context hierarchy (parent) models of a fastWorkflow command tree so intent routing stays accurate as the tree grows and a task takes the fewest commands. Covers deriving the containment graph from code, sizing base contexts to the parameter producers commands actually…
Skill Claude CodeCodex
Read and act on fastWorkflow's near-duplicate capability diagnostics — the pre-flight lexical separability scan and the post-training router-confusion scan — to decide whether two commands should be merged, re-seeded, or recorded as a deliberate near-duplicate. Covers what separability actually measures, the reporting…
Skill Claude CodeCodex
Read fastWorkflow's held-out routing and escalation evaluation honestly, and decide from it whether a change to a workflow's commands, seeds or context models actually helped. Covers why the training F1 measures memorisation, the whole-persona holdout and what it does and does not remove, which of the reported numbers…
Skill Claude CodeCodex
Integrate a fastWorkflow-based agentic chat agent into an existing application so users can execute the app's business logic and ask questions in natural language, with a popup chat UI that streams internal workflow/assistant conversations before the final answer. Use when a developer asks to add a chat agent…
Skill Claude CodeCodex
Implement dblookup on a fastWorkflow command parameter so a value the user typed is resolved against a live key set, typos are corrected automatically and near misses come back as "did you mean" suggestions instead of a not-found. Covers the field declaration, the Signature.dblookup contract, the three-stage fuzzy…
Skill Claude CodeCodex
Decide who writes a fastWorkflow's synthetic training utterances, by adding a personas.json that either supplies the personas outright or conditions the PersonaHub draw on the application's domain keywords. Covers the file schema and where it is discovered, choosing between the two modes, the persona id rules that…