Repository navigation
docs: slim FAVA startup guidance, keep detail in usage guide - #148
Conversation
timeleft--
left a comment
There was a problem hiding this comment.
Native Captain review of 798b54e against issue #146 and epic wise-agents-toolkit#112.
Blocker: integrate the change with current main through a new fast-forward-pushed commit. GitHub reports CONFLICTING; an independent merge-tree probe reproduces conflicts in CHANGELOG.md, server.py, and the server-instruction tests. Current main delegates instructions to mcp_context's full/compact serializers, whereas this old-base change rewrites the earlier inline server implementation. Resolving by restoring that inline implementation would discard the existing compact surface. Implement the accepted concise/on-demand guidance in the current full and compact instruction sources and retain current tool surfaces and protocol behavior. Update affected usage-guide descriptions and test the actual exposed variants.
Blocker: apply the current repository version policy after integrating main. Current main requires an affected-version update for docs/instruction changes as well as code; the PR's stated docs-only exemption comes from its obsolete base and is not the current policy. Use the existing updater and validate against current main.
The intended on-demand wording and preservation of detailed usage guidance are in scope. Installed-server verification remains separately required in the final laptop acceptance ticket. No Captain edits or pushes were made to the worker branch.
…idance to mcp_context full/compact sources
… after merging main
timeleft--
left a comment
There was a problem hiding this comment.
Independent Captain re-review of exact worker head5e54b3fac67cfa04829be3a65ddcc9a4e789c204 in a separate checkout. Prior branch conflict/current-source and version findings are resolved: current main is included, the updater check passes, and all current required checks are green. Captain tests/test_server_instructions.py and tests/test_mcp_context.py pass (61tests), including live MCP/usage-guide seams. The on-demand recall change is correct and tool capabilities remain present.
One remaining acceptance blocker: #146 requires concise startup guidance with detailed procedures moved to the existing usage guide. The actual default full initialize seam still emits4,659characters versus4,734on current main (75characters saved), including the During Work procedure, source_type/type examples, detailed identity examples, lexical-search explanation and manual VCS publication commands (jj bookmark set main -r @- / jj git push --bookmark main). Those details still arrive before get_usage_guide is called, so the PR does not deliver the accepted startup-guidance reduction and its description overstates the slimming. Keep the essential provenance, authority, visibility, scope, on-demand retrieval and safe-operation boundaries in a concise startup entrypoint; retain detailed procedures/examples in AGENTS_USAGE_INSTRUCTIONS.md returned by get_usage_guide. Exercise the actual default and compact initialize seams and retain all17tools and their existing schema/authority contracts. This is the accepted #146 requirement, not a new compact-catalog or broader #54 scope; do not change #54's parent or scope.
Update the context measurements and tests for the final guidance. Also refresh the stale PR body, which still claims no version bump and an older845pass/7failure run; your repair's version bump/full1157pass result should be represented accurately. Keep the same PR/history, no force push. Installed-server/laptop acceptance stays separate and remains pending.
timeleft--
left a comment
There was a problem hiding this comment.
Independent Captain re-review at db44f0f: the startup-guidance finding is resolved. Actual full/default instructions are 2362 characters versus 4734 on main; compact instructions are 836. Detailed procedures and publication commands remain in get_usage_guide. All 17 tool definitions and schemas are unchanged except the accepted on-demand description updates. Captain instruction/context protocol tests: 62 passed; current-main ancestry and owning version-policy check pass. Final measurements reproduced with package 0.8.5rc1 and the reviewed SHA; chars/4 figures are estimates, not measured client billing. Native review/fallback is clean, reusing valid prior evidence. Source readiness is approved; installed-server and laptop acceptance remain pending under #118.
Resolves #146 (parent: wise-agents-toolkit#112; toolkit #54 intent reconciled, not modified).
Changes
fullinitialize instructions cut from 4,734 chars (main) to 2,362 chars. Kept: provenance/authority (governed institutional record), scope discovery order, on-demand retrieval, mandatory promotion, governed visibility, identity authority, recalled-thought safety, lexical recall one-liner. Moved to the usage guide only: During Work procedure,source_typeexamples, detailed identity examples, manualjj bookmark set main -r @-/jj git push --bookmark maincommands.get_usage_guide; on-demand wording retained. All 17 tools and their schemas/authority contracts are unchanged.mcp_contextfull/compact sources (main's structure), not the old inline server implementation.docs/mcp-context-overhead.mdrefreshed: full session-init 6458 tokens (instructions 591), compact 3395 (0.53 of full, budget met).scripts/mw-version.pyper the current version policy; CHANGELOG entry under Unreleased.Verification
uv run pytest: 1158 passed. Version policy check passes against origin/main.Acceptance note
Merged-source acceptance only. Installed MCP servers pick up the change after release and vendor update; the installed-server check stays in the laptop acceptance ticket.