Fireline rulebook

The rules Fireline follows on every job, from the day the drawings arrive to the final report.

What varies: anything that differs between architects, engineers or jurisdictions is recorded in that job's drawing profile (section 8). It is never hard-coded and never assumed.

Service decisions come from the Fireline service plan. Firestop requirements are paraphrased from the IBC sections cited and from the International Firestop Council's training Firestop Basics – Penetrations (PEN2, © IFC 2025; not reproduced here).

The four rules that never change:

1. Fireline never signs off on code inspections.
2. Fireline never changes the drawings.
3. Every status says where it came from.
4. Rated assemblies are determined first, for every set, from that set's own evidence (section 0). Nothing is pinned until the rated-assembly map is approved.

---

0. Determining what is fire-rated (always first, on every set)

Every architect marks ratings differently, and so do different sheets within one set. There is no universal symbol. Each set must be read and reasoned through from scratch. What we learned on an earlier job is a hint about what to look for, never an answer. Only a rated wall or floor needs firestopping, so this step decides everything after it.

0.1 Find every source of rating evidence in this set

Look for all of these. Write down which exist and on which sheets:

0.2 Learn this set's language before reading the plans

• Read every legend that applies: the life-safety key, wall-type legend, partition tag symbol, abbreviations and assembly schedule.
• Record in the drawing profile exactly how this set marks a rated wall, with the sheet and a location you can point to. For example: "1-hr fire barrier = row of solid squares, 'X FB' tag, G-012 key" or "2-hr = gray dash-dot-dot line, a1.04 legend".
• Find one example of each rating on the actual plan and confirm it matches the legend before trusting any symbol.
• Check that no symbol is used for two things. For example, matchlines drawn in the same style as a rated-wall line. Follow every rated-looking line to its ends before believing it.

0.3 Reason out what the code requires (the expected list)

From the code analysis, list every assembly that must be rated, with hours:

• floors and roof, from the construction type;
• shafts and exit stairs: 2 hr at four stories or more, 1 hr below;
• elevator hoistways;
• corridors;
• dwelling-unit separations: walls and floors;
• occupancy separations;
• incidental rooms: boiler/furnace, laundry, trash/compactor, storage over the limits;
• fire walls;
• smoke barriers.

Use the code edition the set names (NYC BC 2022 for NYC). This list is a check, never a substitute for what the architect drew.

0.4 Build the drawn list, one source at a time

• Trace each source independently: life-safety lines, wall-type tags through the schedule, notes, MEP damper evidence.
• For each rated wall or floor record: location, rating (hours), type (fire wall, fire barrier, fire partition, smoke barrier, smoke partition, shaft, horizontal assembly), UL design number if given, full height or to-deck, new or existing, and which sources show it.
• Follow a wall tag all the way: tag on the plan, then the schedule row, then hours and UL design. A tag with no schedule entry is a question, not a guess.

0.5 Reconcile and decide

Things that never prove a rating on their own: wall thickness, CMU, room name, double lines, or a rule from another job.

0.6 Publish the rated-assembly map, then get it approved

• For each floor, the map is a sheet showing every rated wall and floor in color by rating, labeled with hours and source. Open questions are marked.
• The project lead or the GC reviews and approves the map before any pinning. Answers to questions update the map.
• Revisions, ASIs and bulletins re-open the map for the affected areas.
• Once approved, it is the only wall layer the pinning and double-check steps use.

---

1. Intake and scanning rules (getting drawings in)

On a real job the GC sends the drawings at the first audit (setup). Fireline doesn't hunt for drawings; it asks for exactly what it needs, uploads all of it at setup, and logs what arrived. Public sets are only for testing the method.

Which issue to pin from:

First-audit request list (send to the GC before setup; mark each item received or missing):

1. The current CD set, issued for construction, all disciplines, with the date and revision of every sheet: - General: cover, index, code analysis, life-safety / code plans; - Architectural: floor plans, the wall-type legend and partition schedule, floor/ceiling and roof assembly types, sections, firestop and head-of-wall details; - MEP: plumbing (waste/vent and supply plans, risers), mechanical (ductwork and piping plans, damper schedule), fire protection, electrical (power and systems plans).
2. Every ASI, bulletin and revision issued since, and how new ones will reach us.
3. CAD/BIM files (DWG or Revit/IFC), if the architect or engineers will release them.
4. MEP coordination or shop drawings, if they exist.
5. Spec section 07 84 00 (firestopping), and 07 84 43 (joints) if separate.
6. The firestop installer's submittal (listed systems), once approved.

7. The special inspection agency's name and its TR1 scope (NYC).

8. Ask for the best format first. In order:

9. Revit/IFC or DWG;
10. vector PDF exported from CAD;
11. scan.

Ask the GC on every job. Note the format in the profile.

9. Get the whole set for each floor: - cover and index; - code sheet; - life-safety plans; - architectural floor plans; - plumbing, mechanical (duct and piping), fire protection, electrical; - legends and symbol sheets; - firestop details; - damper schedules.

Also get spec section 07 84 00 if available.

10. If we scan paper ourselves (proposed):
    • Resolution: at least 300 dpi, and 400 dpi when the smallest text is under 3/32".
    • Color: grayscale, never black-and-white. Gray background lines carry meaning.
    • Framing: one sheet per page, the full sheet including the title block and scale bar.
    • Clean scan: flattened, no fold shadows, deskewed within 0.5°.
    • Naming: each file by sheet number and revision, e.g. P-202-A_rev3.pdf.
11. Keep the originals untouched. Store them privately as received. Fireline's work always goes in a separate layer or file.

12. Check every set on arrival:

    • Text layer: does the PDF have one? If not, OCR is needed.
    • Scale: is there a scale bar or scale note?
    • Completeness: is any sheet in the index missing?
    • Issue: which issue is it (bid, CD, ASI, bulletin), with its date and revision deltas?

    Record all of it in the profile.

13. Client drawings stay private. They never go to public repos or outside models that retain data. Public test sets are the only exception.

2. Reading rules (understanding the set before touching it)

1. Read in this order:
2. index;
3. code sheet;
4. life-safety plans;
5. symbol and legend sheets (A-001, M-000, P000, FP000);
6. floor plans;
7. MEP plans for each floor;
8. details and schedules.
9. The code sheet sets the scope. Record: - construction type and whether the building is sprinklered; - the hour rating of floors, shafts, exit stairs, fire walls, fire barriers, smoke barriers and corridors; - whether special inspection applies (NYC BC §1705.17: always).
10. Floors count only if they are rated. IA/IB: every slab penetration is an item. IIB: floors are 0-hour, so only shafts, stairs and walls.
11. Rated walls come only from the approved rated-assembly map (section 0). Never judge a rating from wall thickness or room use.
12. Learn this set's symbols from its own legend before reading the plans: rated-wall linetype or symbol, damper tags, riser symbols, abbreviations.
13. Lines that are never pipes: - matchlines; - area or unit boundary lines ("SEE SHEET … FOR CONTINUATION"); - sprinkler zone lines; - the engineer's own fire-wall marker drawn along a rated wall; - grid lines, dimension and leader lines; - title-block and certification lines.
14. Labels: - Follow the leader line to the symbol before trusting a label. - DN or UP/DN goes through this floor. - UP only goes through the next floor. - Waste, soil and storm stacks (CRW, SAN, S, ST, OD, RD) continue down. - A cleanout (CO) is not a penetration.
15. Listed system numbers (UL format) encode the assembly and penetrant: - First letter: C = concrete or masonry floor/wall, F = floor, W = wall. - Number range: 1000s metallic pipe, 2000s nonmetallic pipe, 3000s cables, 4000s cable trays, 5000s insulated pipe, 6000s miscellaneous electrical, 7000s mechanical (ducts), 8000s mixed groups.

Check the submitted system matches the assembly and penetrant the pin records.

9. New vs existing: note what the set marks as new, existing to remain, or NIC. Some sets mark only new barriers. Ask the GC about existing rated walls.
10. Write down every unclear point as a question for the GC. Don't decide it silently.

3. Pinning rules (making the item list)

What every pin records:

1. Exact position: the crossing point itself, not an area.
2. Assembly: wall or floor, its construction (CMU, concrete, gypsum, ...) and its hour rating. Every pin names the rated wall or floor it goes through.
3. Penetrant: pipe, duct, conduit, cable or tray; metallic or nonmetallic; size; insulated or not.
4. Required ratings: - Walls: an F rating of at least the wall rating. A T rating is never required for wall penetrations (IBC 714.4.1.2). - Floors: F and T ratings, each at least 1 hour and at least the floor rating (IBC 714.5.1.2). The T rating is not required for:
   • penetrations contained in a wall cavity;
   • floor, tub and shower drains within a concealed floor/ceiling space;
   • metal conduit of 4" or less into switchgear.
   • Smoke barriers: an L rating as well (IBC 714.5.4).
   • Sprinkler pipe in seismic areas: annular space of 1" for pipe 3.5" or smaller, 2" above that (NFPA 13).
5. The listed system, once the installer submits it: the UL/Intertek number, for example C-AJ-1379.

How candidates are made:

1. Vector sets: - The automatic pass computes crossings from the linework. - On a new architect's set, treat it as candidates only.
2. Scanned sets: - Register the sheets by OCR of the grid bubbles, or by geometry. - Trace the rated walls by hand from the life-safety plan, about 10 minutes per floor. - The visual pass then proposes candidates.
3. Never draw candidates on client-facing sheets.

4. Checking rules (the double-check)

1. Clean pages. Crops of the original MEP sheets with only the rated walls tinted and a lettered cell grid. No pins.
2. Blind re-scan. A separate checker lists every crossing per cell, and judges every slab candidate from its label.
3. Reconcile: - both agree: confirmed on drawings; - the checker saw the spot and found nothing: rejected; - the checker found a crossing with no pin: missed.
4. Tie-break every disagreement with wider crops and a fresh checker.
5. Inferred is not confirmed. If the label was cut off, there's no DN, or it's a judgment call, the item is check in field.
6. A missed penetration is worse than an extra pin. When in doubt, keep it as "check in field".
7. Check the wall layer by eye against the life-safety plan once per floor. A rated wall we failed to extract is invisible to both passes.
8. Keep the removed list with a reason for each item. "UP only" items move to the next floor's list.

5. Tracking rules (field status over the life of the job)

1. Statuses:

2. Every status change records who reported it, when, the source (walk, email, photo, agency report number), and the old and new status. The history is append-only and never edited.
3. Fireline records statuses; it never approves them. The GC signs off "Ready for inspection" and the agency accepts "Complete".
4. Weekly walk: - A Fireline person walks each active floor every week and updates statuses from what they see. - Photos are required for every status set from the walk (proposed). - GC reports between walks are entered the same way, with their source.
5. Field-added penetrations found on the walk get a new pin flagged field-added, with a photo. Items the drawings show but the field doesn't are flagged not found. Never delete them.
6. "Check in field" pins are resolved on the first walk that reaches them: confirmed, moved, or not present.
7. Destructive-test tracking (NYC, ASTM E2174/E2393): - Penetrations: at least 2% of each firestop type (UL/EJ system per installer), with a minimum of 1, per 10,000 sq ft inspection area. - Joints: 1 per joint type per 500 linear feet. - Results: record DT PASS or DT FAIL with the agency report reference. - Failures: a type with 10% or more failing is flagged until the installer has rechecked all of that type.
8. What the walk looks at for each installed item (from the listing, both sides of a wall, top side of a floor): - the installed system matches the submitted listing (assembly, penetrant type and size); - annular space within the listing's range. Note which case it is: centered, off-centered, point contact or continuous contact. Point or continuous contact only where the listing allows it, and point contact needs a crown bead at the contact point; - cable penetrations: percent fill of the opening within the listing; - packing or insulation material and depth as listed; - sealant (fill) depth as listed (listings give wet thickness; some sealants shrink), with a crown bead if listed; - sealant tooled and bonded to both the penetrant and the wall or floor. Untooled sealant sitting on the surface is a likely failure; - floors: T-rating protection (pipe insulation or wrap) where required; - smoke barriers: an L-rated system; - the wall's own stencil above the ceiling ("FIRE AND/OR SMOKE BARRIER – PROTECT ALL OPENINGS"). Note any rated wall without one, and any stencil on a wall the map shows as unrated. Both go back to section 0 as questions.

Record findings as observed; never as approval.

9. Readiness is tracked per floor (headline), then DT coverage per type and area, then failure clusters.
10. Retention: keep records at least 6 years (NYC).

6. Revision rules

1. The architect's sheet is never edited. A revision arrives as a new version of the sheet.
2. Re-pin onto the new revision: - Carry each pin's history over. - Flag pins inside revision clouds, or near changed MEP runs, as check in field until re-read.
3. Every report names the drawing revision each sheet was pinned from.
4. Removed items are archived with a reason, never deleted: - removed by revision; - not found in field; - moved to another floor.

7. Report rules

1. Each issue gets a number (Report 1, 2, 3…), with a date and the drawing revisions it used.
2. Each report contains: - the architect's sheet, untouched, with the Fireline layer on top: pins, rated walls, legend and margin strip; - zoom detail pages (key plan plus enlarged areas with pin tables); - a status table.
3. Targeting: - Each pin sits on the exact crossing point. - Detail pages zoom to roughly 10–15 ft around their pins. - Wall pins and floor pins use clearly different markers.
4. Pin marks: - confirmed on drawings: filled; - check in field: hollow; - field status shown by color.

Never the word "verified".

5. Every page carries: "Pins placed by Fireline from the drawings. Status as observed or reported to Fireline. Not an inspection, certification or code approval."
6. Who receives what: - GC: ongoing reports. - Architect and developer: view as wanted. - Inspection agency: the final report only.
7. Report links are private and expire.

---

8. The drawing profile (one per job, filled in at intake)

Expect every set to be different. The profile records how this set draws things before any reading or pinning. The automatic pass reads its settings from here. Any blank field means "candidates only" for that item.

job: ''
set_issue: '' # "100% CD bid set, 2017-12-16"
format: vector | scanned | mixed
text_layer: yes | no | partial # no -> OCR

code:
  construction_type: '' # IB, IIB, ...
  sprinklered: yes | no
  floor_rating_hr: 0 # 0 -> no slab items
  shafts_hr: 1
  exit_stairs_hr: 1
  fire_walls_hr: 2
  other: '' # smoke barriers, corridors, incidental uses
  special_inspection: yes | no

rated_walls:
  source_sheet: '' # G-012 life-safety plan / a1.04 floor plan
  symbology: '' # "gray dash-dot-dot 1.92 pt" / "chain of 4.6 pt squares"
  only_new_marked: yes | no
  scale: ''

sheets:
  split: whole-floor | zones | units
  scale_ratio_to_rated_sheet: 0.5
  rotation: 0 | 90
  registration: grid-text | grid-ocr | geometry
  background: 'gray 0.67-0.7'
  system_lines: 'black >= 0.45 pt'

ignore_lines:
  - matchline: 'heavy dashed, sheet names at ends'
  - area_boundary: "dashed, 'SEE SHEET ... FOR CONTINUATION'"
  - fire_wall_marker_on_mep: 'dashed black along rated wall'
  - sprinkler_zone_lines: 'heavy dash-dot'
  - title_block_column: 'right ~600 pt'

labels:
  riser: 'UP / DN / UP/DN'
  dampers: 'F, FS, SD'
  stacks: 'CRW, CRV, SAN, V, S, OD, RD'

questions_for_gc: []

9. Job start checklist

1. Request the best format; log the set and issue (section 1).
2. Determine rated assemblies for this set (section 0): find the evidence sources, learn the set's legend language, build the expected list from the code sheet and the drawn list from each source, then reconcile. Fill in code: and rated_walls:.
3. Publish the rated-assembly map and get it approved by the project lead or the GC. Nothing is pinned before this.
4. Open one MEP sheet per discipline; fill in background, system_lines, ignore_lines.
5. Align one sheet and check the overlay by eye.
6. Run candidates (section 3), then the double-check (section 4), then annotate (section 7).
7. Send questions_for_gc with the first report.
8. Log new variance in section 8.