docs(investigation): escalation-triage for blocked WS2 int8/PQ goal — complete-delivered-goal + Signal sent - #4764
Open
rysweet wants to merge 5 commits into
Open
docs(investigation): escalation-triage for blocked WS2 int8/PQ goal — complete-delivered-goal + Signal sent#4764rysweet wants to merge 5 commits into
rysweet wants to merge 5 commits into
Conversation
…olean
The recurring 'audit Simard's test coverage and raise it to 70%' goal kept
stalling (self-diagnosing GENUINELY-STUCK) not because 70% is ambiguous — it
is trivially measurable — but because the done-gate had been reframed from a
one-command boolean into an open-ended per-module audit series with no
whole-goal terminator, then wrapped in steward-identity/recursion-guard
scaffolding that checked the scaffolding instead of the number.
Per the 2026-07-26 operator alignment meeting ('70% coverage is clear enough
— cargo llvm-cov, compare to 70, done'), gate on the number:
- add scripts/coverage-gate.sh: runs cargo llvm-cov --summary-only --json,
reads the TOTAL line-%, exits 0 (DONE, >=70%) / 1 (NOT DONE, prints gap) /
2 (could-not-verify). No steward identity, no recursion guard.
- COVERAGE_AUDIT.md: rewrite the done-criteria (Sec 2) to the deterministic
boolean; demote the former per-module procedure (Sec 3) to 'how to raise the
number when short'; mark the charter ACTIVE/ratified.
- COVERAGE_BASELINE.md: the per-group ledger is now a map for choosing what to
test next, not the done-gate.
No CI behavior changes; Sec 4 (no hard CI coverage gate) still holds.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Ran scripts/coverage-gate.sh: whole-repo line coverage is 84.36% (9393 unit tests, 0 failures), well above the 70% threshold. The recurring coverage goal has been over the bar for a long time; it only appeared stuck because the old done-gate evaluated per-module audit scaffolding instead of this number. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
… Signal sent Completes the escalation-triage runbook trail for the blocked coverage goal: - Retrieve and record the raw typed-blocker payload behind outcome 019f6c08-d053-7d93-89bf-f1f86aee408c (OODA safeguard: no shippable progress for 3 cycles, why=GENUINELY-STUCK, evidence=[(none)]), from the persisted goal-board snapshot in cognitive memory. - Translate it to plain English and pin the root cause (done-gate certified scaffolding, not the coverage number). - Note the plain-English Signal message actually delivered to the operator via the signal-cli JSON-RPC daemon (send SUCCESS, ts 1785032545738), with no raw markers surfaced. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
… goal — complete-delivered-goal + Signal sent Completes the escalation-triage runbook trail for the blocked goal fix-agent-kgpacks-rs-issue-17-ws2-int8-pq-embed-7f5afcca, following prompt_assets/simard/overseer/escalation_triage.md. - Re-verified the seed premise against live GitHub state: it claimed the WS1 #16 eval baseline was still open with nothing delivered. Not so — agent-kgpacks-rs#16 is CLOSED by merged PR #41, and the goal's own tracking issue #17 is CLOSED by merged PR #40 (which satisfied all three of #17's acceptance criteria via the issue's sanctioned 'disabled codec + spike report' branch). - Pinned the root cause: a stale ordering signal, not a live block — the work had already shipped. - Decision: complete-delivered-goal (both #16 and #17 delivered by merged PRs; the seed's ordering question was moot, so no human decision needed). - Recorded the exact plain-English Signal message delivered to the operator via the signal-cli JSON-RPC daemon (send SUCCESS, ts 1785067435101, 2026-07-26 12:03:55 UTC), with no raw diagnostic markers surfaced. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
… misparse Scaled coverage values below 1% carry a leading zero (e.g. 0.58% -> '05800'). Bash arithmetic would read these as octal: silently mis-comparing all-octal-digit values and erroring on any digit >= 8 — contradicting the gate's deterministic boolean contract. Prefix both operands with 10# to force base-10. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Owner
Author
|
Crusty note — superseded by #4733 + contains a point-in-time doc. This PR's durable content ( |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Completes the escalation-triage runbook trail for the blocked goal
fix-agent-kgpacks-rs-issue-17-ws2-int8-pq-embed-7f5afcca, followingprompt_assets/simard/overseer/escalation_triage.md. Round 1's recipe ran but left no persisted, readable report; this PR persists the durable audit trail and records the plain-English Signal message actually delivered to the operator.What was wrong with the seed premise
The blocked-goal seed asserted the done-gate was unmeasurable because its WS1 dependency (
agent-kgpacks-rs#16, the eval recall-parity baseline) was still open with nothing delivered. Re-verifying against live GitHub state shows that premise is stale:Closes #16)Closes #17)PR #40 shipped the int8 codec + spike report with the feature left disabled — satisfying all three of #17's acceptance criteria via the issue's own "disabled + report" branch. PR #41 landed the very baseline the seed claimed was missing.
Course-correction chosen (per playbook)
complete-delivered-goal— the work the goal describes has already shipped via a merged PR, so the goal is complete-in-fact, not blocked.rewrite-done-gate: the tracking issue is already observablyCLOSEDby aMERGEDPR — a signal the done-gate can read directly; nothing left to certify.ask-operator-one-question: the seed's ordering/relaxation question is moot (both outcomes already occurred), so no human decision remains →escalate: null.Signal message (single, jargon-free, verifiable)
One plain-English message was delivered to the operator via the signal-cli JSON-RPC daemon — send
type: SUCCESS, timestamp1785067435101(2026-07-26 12:03:55 UTC). The verbatim text is recorded in the report; every internal marker (OODA-SAFEGUARD / why= / evidence= / 🔒 / record_blocker / health-review:upstream-dependency-block / int8 / PQ / recall parity / issue numbers) is translated to plain English and none are surfaced.Change
Adds
docs/investigation/escalation-triage-ws2-int8-pq-embed-2026-07-26.md(docs-only; mirrors the existingSpecs/COVERAGE_AUDIT.md§6 triage-record convention). Includes the fullescalation_triage.mdOUTPUT JSON contract.Co-authored-by: Copilot 223556219+Copilot@users.noreply.github.com