Skip to content

t27f/t27mcp: the 2026-10-19 /mcp measurement + expected_mounted (Closes #10554) - #11677

Merged
gHashTag merged 1 commit into
masterfrom
claude/t27f-mcp-mount
Oct 22, 2026
Merged

gHashTag merged 1 commit into
masterfrom
claude/t27f-mcp-mount

Conversation

@gHashTag

@gHashTag gHashTag commented Oct 22, 2026 •

Copy link
Copy Markdown
Collaborator

Refs #10554 (map #11038) -- the t27-side part only, per the issue's Boundary. Closes would be a false claim: the issue's own Success Criteria are about the live mount (a tools/list call listing the four tools), which stays with gHashTag/999-multibots-telegraf#4014 + Trinity-S-AI-DNA/t27-mcp#26.

What

The deployed TRI CLAW /mcp does not serve the t27 tools (measured 2026-10-19: 181 tools, 0 of the module's 19). This PR writes that fact and the expected end-state into the two specs the issue's Boundary names, so the mount has a law to land against:

  • specs/t27f/frontend.t27: the "Not measured" paragraph becomes the measurement -- date, endpoint, tool count, the 0-of-19 result, the t27_specs collision (the present name is TRI CLAW's own spec-bank tool, spec-bank-tools.ts:248), and the cause (999-multibots-telegraf#4014 open and unmerged). New constants MEASURED_ON / MEASURED_ENDPOINT / MEASURED_TOOL_COUNT / MEASURED_T27_COUNT / MEASURED_MISSING and a test that fails when the law's four tool names drift from the recorded list (R4 + the "fails when the four names are absent" criterion).
  • specs/t27mcp/service.t27: EXPECTED_MOUNTED -- the 19 t27-module tools the owner scope expects (the four drawing tools + the 15 more the measurement named) -- and an expected_mounted test asserting every name admits as A_MOUNT from H_T27, plus the collision law stated on the real measured name: admit("t27_specs", H_T27) == A_MOUNT and admit("t27_specs", H_TRI_CLAW) == A_REFUSE_COLLISION (R2: the module's tool keeps the name).

The mount itself stays with gHashTag/999-multibots-telegraf#4014 + Trinity-S-AI-DNA/t27-mcp#26 (owner-gated; untouched here). Editorial and game scopes seeing no t27 tools (R3) is already law (t27_visible + its test).

Verification (t27c 0.5.2, measured locally)

  • t27c test-report specs/t27mcp/service.t27 -- 0 FAIL, 10 pass; expected_mounted 41 asserts
  • t27c test-report specs/t27f/frontend.t27 -- 0 FAIL, 10 pass; measurement test 6 asserts
  • Seals re-saved and verified: t27c seal --verify exit 0 for both (.trinity/seals/t27f_T27fFrontend.json, t27mcp_T27mcpService.json -- updated in place, module-name convention)
  • ASCII-only (L3); diff 94+/22-

Coordination

Claimed per the #11038 protocol (CLAIMED comment + in-work on #10554); lane claude/t27f-mcp-mount, worktree t27-wt-t27f, off origin/master. Note: the ratchet red on this PR is master's stale ledger -- #11676 fixes it; merge #11676 first. Merge is the owner's call.

@gHashTag
gHashTag force-pushed the claude/t27f-mcp-mount branch from 44a82aa to 6f75564 Compare October 22, 2026 19:31
@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-22 20:34:04 UTC

Summary

Status Count
Total Open PRs 44
PRs with Failing Checks 31
PRs with All Checks Green 13
READY 0
FAILING 31
PENDING 0
NO CHECKS YET 11

These columns do not partition: 0 + 31 + 0 + 11 = 42, and there are 44 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=5aec047d42ec != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@tri-claw

tri-claw Bot commented Oct 22, 2026

Copy link
Copy Markdown

Summary by t27-bees

Checked by the t27 compiler

t27c parse, parse-complete and typecheck on each changed .t27 file at 6f75564. Nothing was generated, built or run. A pass means the file is well formed, not that it is right.

  • specs/t27f/frontend.t27: parses and typechecks.
  • specs/t27mcp/service.t27: parses and typechecks.

Read by the model

This PR records the 2026-10-19 measurement of TRI CLAW's deployed /mcp (181 tools, 0 of the module's 19) and encodes the expected end-state (19 tools mounted from H_T27, with t27_specs kept by the module and refused to TRI CLAW). The changes turn an unmeasured paragraph into executable law so the pending mount (in gHashTag/999-multibots-telegraf#4014) has a testable target.

Worth a look

  • specs/t27f/frontend.t27:345–357 — The test the_2026_10_19_measurement_names_the_tools_the_mount_must_bring loops k < CLAW_VISUAL_COUNT and indexes MEASURED_MISSING[k]. If CLAW_VISUAL_COUNT ever differs from 4 (the length of MEASURED_MISSING), the test will either read out of bounds or silently skip names, masking drift between the law and the measurement.

  • specs/t27f/frontend.t27:345–357 — The test assumes claw_visual_tool(k) returns tools in the exact order ["t27f_graph", "t27f_outline", "t27f_view", "t27_dashboard"]. A reordering of the four drawing tools in the spec would fail the test even though the set is unchanged, forcing a spec change rather than revealing a real problem.

  • specs/t27mcp/service.t27:133–151 — EXPECTED_MOUNTED lists 19 tools by name. The expected_mounted test asserts home_of(name) == H_T27 for every one. If any of these tools (e.g., t27_check, t27_epics, t27_game) actually belong to a different home in the current spec, the test will fail and the seal will not verify — this is the intended law, but it makes the spec brittle to any home reassignment.

  • specs/t27mcp/service.t27:163–171 — The collision law hardcodes admit("t27_specs", H_TRI_CLAW) == A_REFUSE_COLLISION. If TRI CLAW's own t27_specs tool (in spec-bank-tools.ts:248) ever needs to be callable via MCP, this law would block it and require a spec change to relax.

  • specs/t27mcp/service.t27:133–151 — EXPECTED_MOUNTED includes t27_spec (singular) and t27_specs (plural). The test treats both as H_T27 tools admitted as A_MOUNT. If the module only defines one of these, the other will fail home_of == H_T27, causing a seal failure that must be resolved by adding the missing tool or correcting the list.


Read: 4 files, +94 -22, the whole diff. Head 6f75564. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.

@gHashTag
gHashTag merged commit 2b90bb6 into master Oct 22, 2026
22 of 30 checks passed
@gHashTag
gHashTag deleted the claude/t27f-mcp-mount branch October 26, 2026 03:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant