Gather Construction AI

View saved version →

---
name: gather-construction-ai
description: Analyse Gather construction records and supplied project files for site diaries, commercial reporting, resources, programme analysis and evidence-backed claim drafts. Use for Gather and contractor quantity-surveying workflows.
---

# Gather Construction AI

Read [runtime conventions](shared/openai-runtime.md), then choose the relevant workflow below and read its workflow.md. Load only the workflow and references needed for the request. The 66 named workflows are bundled references, not separately installed skills. References such as `$diary` mean the diary workflow below. Resolve paths relative to the referring file; bare docs/, scripts/ and skills/ paths use this skill root. Live access requires the separately configured Gather MCP connection. Discover available tools and their schemas; do not invent tools or data. Preserve each workflow's evidence requirements, draft status and authorisation boundaries.

Fatigue scoring and monitoring are excluded from this edition. Follow the scope boundary in `shared/openai-runtime.md` before choosing tools.

## Workflow catalogue

- [access](skills/access/workflow.md): Analyses possession and access records — planned vs actual windows, restrictions, timeliness. Triggers: access, possession, site entry, access delays.
- [adjudication-pack](skills/adjudication-pack/workflow.md): Assemble a referral- or response-ready adjudication evidence bundle — issue chronologies, records schedules, programme evidence, quantum refs, witness skeletons. Document assembly support, not legal advice.
- [agents](skills/agents/workflow.md): Find the right Gather specialist for tender, commercial, contract, data quality, programme, claims or workforce questions.
- [asbuilt-programme](skills/asbuilt-programme/workflow.md): Reconstructs the as-built programme — actual start/finish/active/dormant per activity — for EOT/claims (as-planned vs as-built). Chat, Mermaid, HTML Gantt and time-chainage (linear works) outputs. Not forecasts (programme) or current-period output (progress).
- [benchmark](skills/benchmark/workflow.md): Compares performance across projects — labour productivity, resource utilisation, progress rates — to identify best/worst performers, for portfolio trends and senior management reporting.
- [bid-advisor](skills/bid-advisor/workflow.md): Tender lifecycle across five modes — Clause sweep, Pricing, Levelling, Programme, Settlement. Achieved-norms pricing, never a price book. Use for "review this tender", "price this bill", "level these quotes", "tender settlement pack".
- [boq-import](skills/boq-import/workflow.md): Import a bill of quantities (CSV or pasted) with user-confirmed column mapping and section hierarchy into the local tender store the pricing build-up reads. Quantities as billed, never adjusted; totals cross-checked.
- [cashflow](skills/cashflow/workflow.md): Project cash forecast — applications in (certification/payment lags per the contract profile), cost out (ledger run rate plus user-supplied lags), retention movements, net curve. Every timing parameter is profile- or user-stated, never assumed.
- [ce-builder](skills/ce-builder/workflow.md): Build a CE notification (NEC clause 61.3), JCT loss and expense application, or FIDIC claim notice from diary evidence. Notification stage — for the priced clause 62 response use ce-quotation.
- [ce-evidence](skills/ce-evidence/workflow.md): Sweeps every Gather evidence source in one pass for a compensation event, claim, or delay into a record-referenced substantiation pack ready for $ce-builder (notification) or $ce-quotation (pricing).
- [ce-quotation](skills/ce-quotation/workflow.md): Price an NEC clause 62 compensation event quotation — Changes to the Prices, Completion/Key Date delay, Defined Cost assessment. Pricing stage after ce-builder's clause 61.3 notification.
- [ce-radar](skills/ce-radar/workflow.md): Proactive sweep for unnotified compensation-event candidates — half-filled and conflicting records graded, not discarded, cross-checked against ce-register.md, clocked against notification time bars. Candidates for review, never asserted entitlement.
- [commercial-director](skills/commercial-director/workflow.md): Commercial reporting across three modes — Portfolio (cross-project RAG review), Forecast (ETC/EAC outlook for one project), Weekly (Monday QS pack). Say which you want, or "portfolio review", "forecast", "weekly QS pack".
- [commercial-profile](skills/commercial-profile/workflow.md): Capture project rates you control — currency, retention, activity/daywork/plant rates, subcontract orders, prelims allowances — so valuation, dayworks etc. price outputs, not just quantities.
- [commercial-report](skills/commercial-report/workflow.md): Generates a weekly or monthly commercial period report — labour, plant, materials, progress, access and key issues aggregated, for commercial reviews and management summaries. No cost or value data; monetary figures are user-supplied.
- [contract-advisor](skills/contract-advisor/workflow.md): Contract interpretation advisor with NEC, JCT, and FIDIC modes — clause interpretation, notice requirements, entitlement framing, procedure checks, no analytics. Other agents delegate clause questions here rather than loading contract skills themselves.
- [contract-profile](skills/contract-profile/workflow.md): Capture the contract as executed — amended timescales, secondary options, Contract Data, and Z clauses — so contract-touching skills use the contract as amended, not the standard form.
- [correspondence-import](skills/correspondence-import/workflow.md): Import exported emails and notices (.eml, PDF, pasted text) into a local correspondence register — dates, parties, user-confirmed classification and CE links — so notification and reply claims cite evidence, not memory. Nothing applied silently.
- [cost-import](skills/cost-import/workflow.md): Import actual cost from your accounting system (COINS, Eque2, Sage — CSV export) with user-confirmed column and cost-head mapping; writes the local cost ledger the CVR reads. Costs as stated, never invented. Cost side only — value comes from valuation/cvr.
- [cost-to-complete](skills/cost-to-complete/workflow.md): Forecast remaining labour hours and estimate at completion (EAC) from actual Gather productivity. Budget/variance needs a user-supplied planned baseline — Gather holds no cost or budget data.
- [cvr](skills/cvr/workflow.md): Build the monthly CVR — records-backed value at commercial-profile rates vs imported actual cost, margin by head, movement since last CVR, user-supplied accruals and liabilities. Draft for commercial review, never the accounts. Cost side needs cost-import.
- [data-assurance](skills/data-assurance/workflow.md): Data readiness preflight — completeness, consistency, plausibility and traceability across records before delay, claims, commercial, or resource analysis. Pass/Conditional/Fail with blockers.
- [data-quality](skills/data-quality/workflow.md): Assess Gather diary data quality — missing narratives, incomplete records, coverage gaps, double-booked people and duplicate shifts. Add "duplicates" for the targeted double-booking sweep; add "review" for a record-level punch-list of defective records.
- [dayworks](skills/dayworks/workflow.md): Build a signature-ready daywork (T&M) sheet — labour, plant, materials on an instructed task, priced at daywork rates. Not subbie verification (use subcontractor-check) or plain labour/plant analysis (use labour/plant-report).
- [delay-analysis](skills/delay-analysis/workflow.md): Builds a delay register from site records — identifies, categorises and quantifies delays from narratives, access restrictions and weather data, for delay claims and lost-time analysis.
- [delivery-notes](skills/delivery-notes/workflow.md): Track and analyse material deliveries — search records, delivery frequency/patterns by supplier or material, reconcile against usage. Not usage/consumption analysis alone (use resource-analysis).
- [diary](skills/diary/workflow.md): Search, list and summarise diary shift records — shifts, what happened on site, supervisor records. Not keyword/theme searches across narratives (use narrative-search) or a single known record ID (use record).
- [estimator](skills/estimator/workflow.md): Builds a labour-hour rate card — unit rates (hrs/m, hrs/m2, hrs/m3, hrs/nr) from output quantities vs labour hours, for productivity benchmarking and effort forecasting. Hours only; monetary pricing needs user-supplied cost rates.
- [events](skills/events/workflow.md): Identifies potential contractual events — compensation events (NEC), Relevant Events (JCT), FIDIC claims — from site diary data. Triggers: find potential CEs, event register, claims significance.
- [fidic-contracts](skills/fidic-contracts/workflow.md): FIDIC Red/Yellow/Silver 2017 and 1999 contract knowledge — claims under 20.2, the 28-day notice bar, DAAB disputes, and variations. Not NEC or JCT (own skills).
- [final-account](skills/final-account/workflow.md): Assemble the running, always-current final account — measured work, agreed vs unagreed CEs with evidence grades, records-supported vs claimed totals, and a negotiation appendix. Internal commercial-review draft, never for issue to the other party.
- [jct-contracts](skills/jct-contracts/workflow.md): JCT SBC/DB/ICD/MW 2016 and 2024 contract knowledge — Relevant Events/Matters, EOT, loss and expense (L&E), LADs, and interim payments. Not NEC or FIDIC (own skills).
- [labour](skills/labour/workflow.md): Analyse labour hours by person, trade, period — who worked when, overtime, utilisation. For roster/headcount (who's on site, mobilisation), use people.
- [material-reconciliation](skills/material-reconciliation/workflow.md): Reconcile delivered material quantities against placed/measured quantities to expose waste, loss, and record errors — delivered vs placed, concrete recon, waste tracing.
- [measured-mile](skills/measured-mile/workflow.md): Calculates productivity disruption using the measured mile method — compares baseline vs impacted period output rates to quantify loss, for disruption and productivity claims.
- [narrative-search](skills/narrative-search/workflow.md): Search site diary narratives for keywords or themes — find evidence for claims across all shift records, mentions of events, issues, instructions, or conditions.
- [nec-contracts](skills/nec-contracts/workflow.md): NEC3/NEC4 ECC contract knowledge — compensation events (CE), early warnings (EWN), the 61.3 time bar, the Accepted Programme, Options A-F, and Z clauses. Not JCT or FIDIC (own skills).
- [nec-deadlines](skills/nec-deadlines/workflow.md): NEC deadline clock — clause 61.3 time bar, deemed acceptance, 61.4 PM reply, 62.3/62.6 quotation, 64/50.9/53 deadlines, from user-supplied dates. "Have we missed the 8 weeks"; time bar; deadline.
- [people](skills/people/workflow.md): Show workforce roster, headcount trends, mobilisation/demobilisation — who's on site, when they started/stopped, staffing levels. Not hours/effort/overtime (use labour).
- [photos](skills/photos/workflow.md): Build a chronological photo/evidence timeline from shift records — dates, locations, supervisor attribution, for progress review or claims. Invoke explicitly (`$photos`); not model-invoked — expensive per-record fan-out.
- [plant-offhire](skills/plant-offhire/workflow.md): Rank idle hired plant by weekly cost exposure and recommend off-hires, for off-hire reviews, idle plant, or hire costing money.
- [plant-report](skills/plant-report/workflow.md): Produce a formal plant status report — register, movements, utilisation, off-hire recommendations — or an Off-Hire Mode ranking idle hired plant by weekly cost exposure. Explicitly invoked only (`$plant-report` or `$plant-offhire`).
- [prelims-monitor](skills/prelims-monitor/workflow.md): Track prelims — weekly staff on site vs tender allowance, elapsed vs tender duration, and a projected overrun at current run-rate.
- [programme](skills/programme/workflow.md): Explicitly-invoked programme forecast — schedule-risk ranking, run-rate and earned-schedule (SPI(t)) forecast completion lenses. No critical path (no logic-link data from API). Not current-period output (use progress).
- [programme-advisor](skills/programme-advisor/workflow.md): Programme artefact specialist — health-checks the Accepted Programme, compares baseline vs actuals, compiles the monthly update pack, reviews a revision, and assembles the 32.2 submission pack. Reads float/criticality as stored, never recomputes.
- [programme-compare](skills/programme-compare/workflow.md): Rev-to-Rev comparison of imported programme revisions — date and duration movement, logic changes, float and TRA erosion per task, all read as stored from each revision (a diff of two already-computed schedules, no recompute).
- [programme-create](skills/programme-create/workflow.md): Draft a construction programme from scope and your achieved norms — user-confirmed logic, engine-computed dates, NEC 31.2 checklist, exported as MS Project XML for the planner to own. A skeleton, never a finished plan.
- [programme-impact](skills/programme-impact/workflow.md): Indicative what-if on the imported programme: add a delay or change a duration and the deterministic CPM engine shows the movement of planned Completion and the critical path, always INDICATIVE-bannered with the self-check % — never the Accepted Programme.
- [programme-import](skills/programme-import/workflow.md): Import the programme from MS Project, Primavera P6 (XER), or Asta Powerproject via the deterministic parser script into a JSON sidecar — baseline tasks, float and critical path as stored, an engine self-check %, and versioned revision history.
- [progress](skills/progress/workflow.md): Tracks current-period progress — planned vs actual output, cumulative S-curves, earned-schedule SPI(t), productivity trends, run rate. Not forecasting (use programme) or as-built reconstruction (use asbuilt-programme).
- [progress-report](skills/progress-report/workflow.md): Generates a formal progress report for a period — progress metrics, narrative evidence, workforce summary, look-ahead — for client or management submission. Explicit invocation only ($progress-report); ad-hoc questions use progress.
- [project](skills/project/workflow.md): Show project overview — team, status, recent activity, who's assigned. Not for financial or cost queries (API exposes no contract sums or rates).
- [quote-levelling](skills/quote-levelling/workflow.md): compare subcontractor quotes like-for-like against the imported BoQ package scope — INCLUDED/EXCLUDED/QUALIFIED/SILENT verdict per scope item, exclusions surfaced, adjusted comparison only with your plug rates; SILENT is a finding, never assumed included.
- [rate-library](skills/rate-library/workflow.md): Portfolio labour-norms library — sweeps recorded output rates (hrs/m, hrs/m2, hrs/m3, hrs/nr) across your projects into a versioned local library with per-norm provenance, compared against user-supplied tender norms. Hours only, no money.
- [record](skills/record/workflow.md): Look up a specific shift record by ID — displays workers, weather, narratives, photos. Use when the user supplies or has just been shown a shift record ID.
- [red-team](skills/red-team/workflow.md): Attacks a draft CE, claim, EOT or final account before submission — verdict tier, opposing-PM letter, ranked fixes. Use for "red team this", "attack this before it goes out", or "sparring mode" (training). Internal war-gaming, never sent to the other party.
- [resource-analysis](skills/resource-analysis/workflow.md): Analyse plant equipment and materials — hire periods, idle time, consumption, planned-vs-actual allocation variance. Not labour (use labour); for a formal plant status document use plant-report.
- [routines](skills/routines/workflow.md): Set up recurring Gather checks and reports from the bundled routine templates when the user requests a schedule or automation.
- [setup](skills/setup/workflow.md): Verify the Gather plugin works — checks MCP connection/auth, lists projects, reports configured dashboard reports. Use when starting out, after install, or a skill reports missing data/tools.
- [subcontractor-check](skills/subcontractor-check/workflow.md): Verify a subbie's claimed hours, output and daywork lines against recorded attendance and deliveries; draft a factual response. Not the main contractor's own valuation (use valuation).
- [tender-pricing](skills/tender-pricing/workflow.md): build the tender pricing from the imported BoQ — labour from your achieved norms, plant/materials/subcontract from your quotes, OH&P as your stated decision; every line labelled with its source, gaps flagged UNPRICED, never a market rate.
- [time-and-claims-advisor](skills/time-and-claims-advisor/workflow.md): Delay and claims specialist across three modes — Diagnosis, EOT, Claim pack — for NEC/JCT/FIDIC. Invoke for delay disputes, extension-of-time assessments, or evidence-backed claims. Clause interpretation delegates to contract-advisor.
- [timesheet-check](skills/timesheet-check/workflow.md): Validate paid timesheets (person, date, hours) against Gather's signing-in attendance and labour hours, verdict per line, both directions. Not a subbie's quantity/daywork claim (use subcontractor-check).
- [valuation](skills/valuation/workflow.md): Build the monthly AfP / interim application, priced at commercial-profile rates, retention and cumulative arithmetic. A draft pack for review, never a certified valuation. Not subcontractor claims (use subcontractor-check).
- [weather-impact](skills/weather-impact/workflow.md): Quantifies weather impact on working time and productivity — lost days from shift weather records and programme impact. Triggers: weather delay, weather claim, seasonal impact.
- [workforce-and-performance-advisor](skills/workforce-and-performance-advisor/workflow.md): Two modes — Forecast (mobilisation/resource planning) and Performance (subcontractor attendance, productivity, programme compliance). Use for manning forecasts, labour shortfalls, subcontractor performance, or recovery actions.

Referenced files: 215 — file references are metadata, not package contents.