Provenance Record

Well-structured News piece (568 words, within the 400-1200 range) using the standard Overview / What We Know / Why It Matters / What We Don't Know framing. Clear, precise, and appropriately technical. The 'What We Don't Know' section responsibly flags undisclosed contract value, electricity price, unnamed renewable projects, and unclear phasing of the remaining 2.7 GW.

Read all three captured snapshots from disk (source-3 failed). source-0.html.gz (RenewEconomy): CONFIRMED the 12-year/600 MW deal with Gunvor, the 1.2 GW / 2032 commitment, the 'more than half of Firmus' initial firming capacity' framing, the 220-hours-per-year demand-response commitment above agreed price thresholds, and verbatim the Oliver Curtis co-CEO quote and the Peter Malinauskas premier quote; it also carries the article's 'Firmus said on Tuesday' attribution verbatim (the deal was announced Tuesday, reports published Wed 1 July). source-1.html.gz (Energy-Storage.news): CONFIRMED 600MW firm, 12-year, 1.2GW, 1.5GWh by 2032, 2.7GW planned across Tailem Bend and Stirling North, Project Southgate, and the 200MW/800MWh grid-forming Koolunga BESS near Brinkworth. source-2.html.gz (pv magazine Australia): CONFIRMED the 200 MW/800 MWh Koolunga BESS, early-2028 operation date, near Brinkworth, and verbatim the Roby Camagong quote (source renders his title 'co-Chief Executive Office'; article uses the standard 'co-Chief Executive Officer'). source-3 (smbtech.au) returned HTTP 403 and could not be snapshotted; it is a supplementary citation and no unique claim in the article depends on it, so its unavailability does not affect verification.

Every specific (600 MW, 12-year term, 1.2 GW renewables, 1.5 GWh storage, 2032 deadline, 2.7 GW planned capacity, 200 MW/800 MWh Koolunga BESS, Brinkworth, Tailem Bend, Stirling North, Project Southgate, 220 demand-response hours, early-2028 operation) traces to a cited source. All three direct quotes appear verbatim in their attributed snapshots. No hallucinations detected.

APPROVE. Integrity checks pass, signature valid, and every specific, quote, and lead claim is corroborated by a source snapshot I read personally. The two automated warnings are editorial-judgment items (off-allowlist reputable outlets; one redundant bot-blocked source) with no impact on the accuracy of any published statement — there is nothing factually incorrect that a corrections note would honestly document, so a clean APPROVE is the correct verdict rather than APPROVE_WITH_CORRECTIONS.