plat 213

← Pulse platform issue index

PLAT-213 — harness:record_pulse_result (ICICI-BANK-PARSING): already fixed by PLAT-206, same code path, different trigger

Coordination	Value
Assigned agent	unassigned
Ticket state	resolved — correction/closure record only, no new code; already fixed by PLAT-206
Last synchronized	2026-08-29

• Priority: P3 — no live gap. Confirms an already-shipped fix covers a second, independently-filed instance of the same defect, on a different workflow.
• Owner: N/A — no code change made or needed by this ticket.
• Related: PLAT-206 — the fix this confirms already covers this finding. harness:record_pulse_result (Workflow/ICICI-BANK-PARSING, high, first seen 2026-08-11) is the finding this closes.

The finding

A live Pulse pass on Workflow/ICICI-BANK-PARSING called record_pulse_result(module=workflow_review, result=done, reason=probe) as a throwaway/exploratory first call — it succeeded. Minutes later, in the same conversation, a second call with a complete, correct payload (changed_files, verification, 8 real finding_dispositions) was rejected: "Pulse module 'workflow_review' for run '...' is already terminal or belongs to another run." The real review outcome could never be recorded; only the throwaway done/probe text stuck as the module's terminal reason for that run.

Same mechanism PLAT-206 already fixed, different trigger

PLAT-206 (filed against a confida-login split reviewer/Fixer finding) fixed markPulseModuleResultFromAgentWithAuditAndFindings's retry guard, which previously only accepted a second record_pulse_result call for an already-terminal (module, pulse_run_id) when its result string exactly matched the first call's. This finding is the identical code path hit by a different real-world trigger: not a separately-dispatched Fixer, but the same session correcting its own premature call. The shapes are identical either way — a first call (result=done, no real evidence) and a second, genuinely different call (result=changed, real changed_files/verification/finding_dispositions) — which is exactly what PLAT-206's relaxed guard (existing.LastResult == result || len(dispositions) > 0) accepts.

Verified directly rather than assumed: wrote a scratch test reproducing this finding's exact reported shape (probe done call, then a changed call with 8-equivalent real dispositions) against the current code — passed cleanly. Not committed as a permanent test since PLAT-206's own TestRecordPulseResultAcceptsFixerSupplementalDispositionsAfterReviewerTerminal already exercises the identical code path and would regress identically if this behavior ever broke again.

Explicitly not done

• No code change — PLAT-206, already shipped, covers this.
• Did not add a workflow-specific (ICICI-BANK-PARSING) permanent test — the mechanism is workflow-agnostic and PLAT-206's existing test already pins it.

Verification

• Scratch reproduction of this finding's exact reported call sequence passed against current code (go test, then deleted — not a permanent addition, see above).
• go build ./... clean; no files changed by this ticket.