Implement features with tests and quality gates, debug hard bugs, and stress-test designs. Use when building a feature with TDD/EDD, chasing a stubborn multi-function bug (feedback-loop-first), interrogating a design before coding, or running a…
Plan-execute-reflect loop over a persistent Python kernel (the `code-interpreter` MCP): write Python, execute, observe the returned trace (stdout / exception / last value), then revise. Variables, imports, and dataframes survive across every tool call in the…
Fires as a post-task hook after a non-trivial completed task (3+ tool calls, observable terminal outcome) to distil 0-N generalisable IF/THEN lessons from the trajectory and store them in RuVector so they surface at the start of similar future tasks; also…
Unified dispatcher for the full skills estate. Use when you don't know which skill to invoke — describe your task and get routed to the optimal skill, agent composition, or MCP tool. Not for executing the task itself: it only routes, then hands off to the…
Execution-gated branching code generation (ADR-020 Surface 2). Generate N candidate solutions (default ≤5) by invoking `sparc:coder` with varied temperature/framing, execute each in a fresh `KernelSession` via the `code-interpreter` (ADR-018 kernel) MCP,…
Store and retrieve verified, executable Python skill primitives (a function plus assertions plus at least one example) in the code-harness procedural memory tier. Use when a reusable utility, parser, validator, or algorithm has proven general-purpose across…
Trigger on "/podcast-ingest", "weekly podcast ingest", "process new podcast episodes into the ontology", or setting up/debugging the podcast-cron schedule. Weekly cron that downloads new episodes from configured YouTube podcasts, extracts evidence-backed…
Audit and de-slop prose for AI writing tells (lexical, structural, and narrative), strip technical AI provenance marks (invisible Unicode watermarks, statistical sampling fingerprints, C2PA/EXIF/XMP metadata), and enforce UK English. Use when writing or…