Aero123421
UserChain AI coding CLIs (Codex, Claude Code, opencode, Grok, Antigravity) into YAML-defined multi-stage flows — review loops, parallel fan-out, session resume. Single binary.
Categories
Indexed Skills (9)
sfh-ci-monitoring
Design SimpleFlowHarness YAML for triggering, locating, watching, diagnosing, and reacting to CI runs such as GitHub Actions. Use when a flow polls CI, waits for checks, repairs failures, downloads logs or artifacts, handles flaky/infrastructure failures, or must avoid monitoring the wrong run, looping forever, or blindly rerunning broken CI.
sfh-context-workspace
Design SimpleFlowHarness workspaces, named context, profiles, sessions, and execution inputs. Use when deciding current versus managed worktree execution, preventing folder/worktree explosion, passing task or review artifacts to specific phases, separating reusable roles from models, preserving context across long loops, or making resume behavior reproducible.
sfh-deterministic-gates
Design deterministic gates and routing in SimpleFlowHarness. Use when tests, builds, validators, CI conclusions, exact votes, exit codes, stderr patterns, outcomes, labels, or AI verdicts decide what happens next; when separating deterministic facts from nondeterministic model output; or when a flow retries, branches, or gets stuck for the wrong reason.
sfh-failure-recovery
Design failure handling and recovery for SimpleFlowHarness flows. Use when choosing retry, retry_on, fallback, on_error, replay.unfinished, resume, force-resume, adopt-workspace, carry-budget-from, budget landing, timeout, stuck, or idempotency behavior; especially for external APIs, deployments, migrations, CI, and long-running AI work.
sfh-flow-design
Design, write, or refactor SimpleFlowHarness (sfh) YAML workflows. Use when turning a task, engineering process, review loop, CI process, research process, or tool chain into a stable sfh flow; when choosing steps, routes, workspaces, contexts, profiles, retries, budgets, and terminal states; or when an AI-authored sfh flow is vague, brittle, or uses invented syntax.
sfh-flow-review
Audit an existing SimpleFlowHarness YAML flow for correctness, stability, convergence, recovery, security boundaries, portability, context/workspace use, external-tool identity, CI monitoring, and AI usability. Use when reviewing an AI-authored flow, diagnosing a brittle loop, checking a flow before release, or producing prioritized fixes without running paid or mutating steps.
sfh-loop-engineering
Design stable, bounded SimpleFlowHarness loops: review/fix cycles, planner-builder-evaluator systems, long-running chunked work, multi-agent councils, and iterative improvement. Use when a route revisits a step, an AI must improve work over several laps, a run may last hours, or an existing sfh loop oscillates, repeats work, loses context, or never converges.
sfh-tool-integration
Integrate external CLIs, web-search tools, HTTP APIs, repository tools, and MCP-enabled AI agents into SimpleFlowHarness YAML. Use when a flow fetches changing remote information, invokes `gh`, `curl`, search CLIs, or provider MCP tools, crosses a network or authorization boundary, or needs reliable artifacts, rate-limit handling, least privilege, and safe replay semantics.
sfh-eval-engineering
Turn failures, regressions, flaky behavior, weak prompts, and unsuccessful sfh runs into reusable evaluations and better harnesses. Use when improving a flow through evidence, creating regression-first repair loops, separating quality and regression suites, promoting repeated review feedback into deterministic checks or repository guidance, or diagnosing why an agent loop keeps failing.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.