Runtime assemble was stamping Snapshot.SourceHash with the registry
surface hash, so schema --all catalog_hash diverged from the CI dump
content source_hash and failed Policy. Also remap ContractFinal
Interface.Ref onto pinned MCP metadata so interface_type stays aligned.
Co-authored-by: Cursor <cursoragent@cursor.com>
Require param_aliases generate plus assembly determinism instead of a
committed cmd_schema_catalog go:generate path; gofmt and temp cleanup.
Co-authored-by: Cursor <cursoragent@cursor.com>
Keep declare→ResolveSchemaBuild as the ToolSpec authority, but materialize
map[cli_path]CommandMeta during deliverySchemaCatalog sync.Once so leaf
--help / ResolveMeta are O(1) after the first Schema touch. Defer wire
Catalog/Tools maps, stamp Source as runtime-assembled, drop committed
catalog/gob fixtures, and cover steady-state reuse with app/cli tests.
Co-authored-by: Cursor <cursoragent@cursor.com>
Move schema authority to declare-time ParamDecl/ContractFinal and
ResolveSchemaBuild so CI/runtime assemble instead of shipping JSON
exclusions/meta-index as delivery sources.
Co-authored-by: Cursor <cursoragent@cursor.com>
Move AnnotateRuntime* into cli/runtimeannotate and the Cobra-keyed
ContractFinal store into cli/contractfinal so corecmd depends on thin
subpackages instead of the cli delivery root. Keep contract as DTO-only,
document Description declare-vs-delivery, and house homology gates under
cli/homology.
Co-authored-by: Cursor <cursoragent@cursor.com>
Publish a compact schema_meta_index.json beside the catalog so help/selection
lookups avoid decoding the full ToolSpec wire on the hot path.
Co-authored-by: Cursor <cursoragent@cursor.com>
Drop the dual-entry thin alias layer so helpers/shortcut/framework author
contract.* types directly; keep only AnnotateRuntime* delivery helpers in cli
and document the corecmd→cli seam.
Co-authored-by: Cursor <cursoragent@cursor.com>
Remove schema_hints as a generation input so Catalog delivery depends solely on leaf ContractFinal and ProductDecl; migrate policy and contract tests to embedded catalog introspection and fix publicShortcutCount for chat-list.
Co-authored-by: Cursor <cursoragent@cursor.com>
Move the last hint parameter overlays into in-code ParamDecls (helpers +
shortcuts), regenerate catalog, and harden the migrate script for factory
and Use/RPC matching so all 74 overlay tools are declaration-backed.
Co-authored-by: Cursor <cursoragent@cursor.com>
Move 121/210 parameter-level Schema field overlays from
schema_hints/metadata JSON into in-code ParamDecl declarations on
DeclareLeafMetadata commands. The declared values are emitted as
dws.schema.* annotations at assembly time via ApplyParamDecls,
outranking tool_schema_hint (rank 620 > 500) so the hint overlay
becomes redundant once the declaration is in place.
Key mechanism changes:
- Add SchemaDecl.Parameters []ParamDecl with property/required/
interface_type/description/required_when/enum fields
- Add cli.ParamDecl type carried inside ContractFinalPayload
- ApplyParamDecls emits dws.schema.* annotations from the payload
at assembly time (no sync.Map, no tree-rebuild key issue)
- Add cli.AnnotateRuntimeFlagInterfaceType (41 overlays needed it)
- Add cli.AnnotateRuntimeFlagRequiredValue for explicit true/false
- Compatibility alias check tolerates dws.schema.* annotation
differences between primary and alias commands
- Remove runtime source_hash recomputation (87ms/997k allocs saved;
enforced by check-generated-drift.sh at build time instead)
- Add shortcut.Flag.RequiredWhen and wire through FromShortcut
- Add boolFlag OR semantics to fix confirmationBypass disagreement
- Add bindKey default kebab-to-camel for forgotten Bind
- Add schema consumption benchmarks (catalog decode 1.5s/817MB,
shortcut load 1.9ms/3MB — three orders of magnitude apart)
- Add catalog codegen feasibility probe (34 tools: 0.06s compile,
31ns access, 87KB linked — extrapolates to 2.1MB for 845 tools)
Migrated products (29 tools, 121 fields):
aisearch(1), chat(12), contact(4), doc(3), drive(1),
hrbrain(10), mail(2), report(1), sheet(3), todo(6)
Remaining 89 fields across 11 products blocked by:
- RPCName not found as string literal (19 tools, shortcut/variable)
- No matching DeclareLeafMetadata near callMCPTool (11 tools)
- No single RPCName for multi-step commands (drive.upload etc.)
All tests green: corecmd, cli, helpers. Generation and drift clean.
The ContractFinal assembly path dropped every non-declared parameter fact,
breaking the published Schema against the reviewed merge-base contract:
- merge pinned MCP parameter metadata and the reviewed in-code runtime hints
back into contract_final parameter resolution (318 interface_type losses,
calendar recurrence required/required_when regressions)
- restore devapp write risk to high and publish back to write/high; the
migration silently downgraded 14 dev write tools and re-classified publish
as destructive
- keep dev at-least-one checks as Validate hooks with the shipped wording
instead of publishing new typed constraints; constraint publication is a
contract change that belongs to its own reviewed PR (aitable annotations
reverted for the same reason)
- align RunE escape hatch, BoolFlag shadowing, guard-first ConfirmFirst
declaration, and Sheet target preflight with behavioral tests; drop the
retired gen_schema_decls.py helper and fix corecmd naming in docs/CHANGELOG
check-authoritative-schema-compatibility vs origin/main: ok.
Compile reviewed Agent Schema into bind-time Go declarations so catalog
tools stamp contract_final without changing execution bodies.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add the ⑥ regression gate that replays every reviewed validation_fixture bad case
through the real embedded PreParse pipeline and asserts the canonical outcome
(accepting either semantic rewrite or native real-flag acceptance, failing only
on a genuine unknown-flag hallucination). Add check-param-concepts.sh (dictionary
schema/loader invariants) and check-param-alias-cooccurrence.sh (full-tree
co-occurrence scan), and wire all three into make policy.
Add internal/generator/cmd_param_aliases: reads the reviewed dictionary plus the
live Cobra tree, reduces each concept against a command's real flags (>=2 visible
real flags without a reviewed ambiguous entry fails generation), and emits the
committed internal/cli/param_aliases_generated.go table with lookup helpers.
Extend generate-schema and check-generated-drift.sh to treat the dictionary as a
reviewed input and byte-guard the generated table.
matched_tools claimed every surface tool matched even when entries were
registry stubs with no live MCP metadata, and unmatched_tools was
hardcoded to 0. Coverage now excludes stubs from matched_tools, reports
them as unmatched, and a registry JSON parse failure warns instead of
silently producing a stub-only snapshot. The schema catalog policy
invariant is relaxed to match the honest accounting.
Restore capabilities now supported on main (doc read --scope/--tags,
drive upload --node overwrite, chat category, dingtalk-markdown routing),
remove commands still absent from the open-source CLI (calendar event
instances, sheet info --include, chat group create --owner), remap
folded services (attendance/ding/oa/report/sheet) to dingtalk-misc in
the shortcut generator, and regenerate shortcut sections and schema
metadata.
A changed production Go file with no function bodies (pragma carriers such
as internal/cli/gen.go, doc-only files) can never appear in a coverage
profile, so the missing-profile check failed every PR touching one. Parse
changed files and exempt those without executable statements; unreadable
or unparsable files stay conservative.
Publishes all 210 public built-in shortcuts as reviewed Agent-visible
leaf tools across 16 product groups, with stable canonical identities,
executable +shortcut CLI paths, parameter and cross-parameter
constraints, selection guidance, interface metadata, and runtime-aligned
safety/confirmation semantics. Catalog grows from 603 to 813 tools.
Gitee API returns HTTP 200 with null body when a release tag doesn't
exist, unlike GitHub which returns 404. Treat empty release_id as
"no release exists" instead of erroring out.
Several read shortcuts returned an empty list with exit 0 and no error
envelope even though the underlying MCP tool returned data, so agents misread
"no data" and made wrong decisions.
Root causes:
- Container key mismatch: the resolver probed the wrong key —
processCodeList / values / wikiSpaces / itemList / groupList / recentItems /
emailAccounts / deptUserList / labelUserList / roles / report_list, plus
get_org_labels grouped labels[] needing a descend.
- Item fields nested under a VO wrapper, not unwrapped: shiftVO / entityVO /
userInfo.
- Param exceeded a backend limit: todo +created-todos sent pageSize=50 while
the backend silently returns empty for pageSize>20; now uses the shared pager
(pageSize=20).
Affected: contact/oa/wiki/drive/minutes/calendar/attendance/chat/report/smart
resolvers. Every fix ships a guard test that feeds the real backend response
shape (and, for minutes, both the taskUuid and minutesId item shapes) and
asserts the projection is non-empty with a usable id.
scripts/shortcut_real_result.py now compares the upper (projection) output
against the lower (raw backend) layer, and record_real_shortcut_run.py captures
the lower layer in memory (persisting only derived counts, never raw PII) so an
exit-0 empty projection over a non-empty backend is scored as
projection-data-loss instead of real-ok. The Python self-test runs in CI via
test/scripts.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
