Export `transcript.md` + `brief.md` from the current chat as an evidence package for framework improvement, preserving user-authored request text verbatim and capturing the artifacts, decisions, friction, and execution evidence needed for postmortems.
Pre-flight git hygiene before commits and before reporting completion: resolve untracked task-created files (especially `*/src/**`), stage renames/moves, and verify that the staged set matches the intended commit scope.
Improve an AI agent framework from implementation evidence: analyze the model-produced result commit, human fix-up commits, and optional implementation chat evidence; then patch the correct framework artifacts and/or <project-local>/ so a rerun with the same prompt/model no longer needs the same manual fixes.
Improve the AI framework from design-stage evidence: analyze exported chat artifacts plus design documents such as the formal task statement, HLD, and execution spec; then patch the correct framework artifacts and/or <project-local>/ so similar future design chats require fewer clarifications, less rewriting, and produce more executable specs.
Synthesis of an operational description of a new or modified framework concept based on sources (books/articles/methodologies/industry): extract invariants, variants, and anti-patterns, then produce a Concept spec plus a separate 'Sources and Attributions' section to add or revise materials in `concepts/`.
Refactor Spring/JUnit HTTP tests to typed `*HttpApi` clients (controller-typed signatures, `*ForResponse`, optional `*ForError`/`*ForOutcome`, and JSON schema verification inside the client) when migrating away from direct WebTestClient/RestTestClient usage.
Initialization of an engineering-log task directory based on a template, including file creation, copying 04-execution-spec.md, and writing the original problem statement to 01-problem-statement.md.
Turn an informal development task statement into a formal task statement from the user’s point of view, including goal/outcome, context, assumptions, glossary, scenarios, constraints, in/out scope, and acceptance criteria as black-box test cases. Use when requirements are unclear (notes, chat, ticket text) and need to be made testable before solution design or implementation, and when preparing an artifact to hand off to the designer/developer roles.