steward-itlisted

Retrospectively investigates an engineering session when work was unexpectedly slow, difficult, repeatedly off course, or exposed a recurring agent-workflow problem. Reconstruct the session from available conversation, tool, repository, skill-trace, context-cost, and outcome evidence; distinguish observed facts from inference; classify the likely cause across methodology, skill, project, stack, prompt, execution, or external limitation; detect repeated patterns; and recommend deliberate improvements without changing canonical guidance automatically. Invoke explicitly for requests such as 'steward this session', 'what happened here?', 'why did this take so long?', or 'is this a recurring problem?'. Not part of the normal lifecycle and not an automatic reviewer of ordinary work.
elieandraos/agentic-engineering · ★ 0 · AI & Automation · score 70

Install: claude install-skill elieandraos/agentic-engineering

# steward-it ## What this skill does `steward-it` is the retrospective diagnostic companion for Agentic Engineering work that was unexpectedly slow, confusing, wasteful, repeatedly off course, or useful as evidence for improving the system. It investigates the session after the work happened. It is human-invoked only — do not activate it merely because a session is long, a tool failed, or an implementation needed recovery — does not silently monitor ordinary sessions, and does not automatically modify skills, project instructions, or methodology. The goal is: > What happened, why did it happen, is it a one-off or a pattern, and what deliberate improvement is justified? ## Invocation Use when the human explicitly asks for stewardship, retrospective diagnosis, or investigation of the engineering session itself. Typical prompts: ```text "Steward this session." "What happened here?" "Why did this take so long?" "Is this a recurring workflow problem?" ``` ## Workflow 1. Identify the target session and reconstruct the observed execution. 2. Load the conditional rule(s) required by the available evidence and requested analysis. 3. Compare expected and observed behavior against the owning skill contract and project state. 4. Classify supported findings and check for recurrence. 5. Produce the standard compact stewardship report, including the recommendation and, when applicable, a durable evidence record. ## Rules - `session-reconstruction.md` — identifies the target,