Computer-Use Agent Taxonomy中文
R08.12.01.01 · TASK CLASS

ELN, LIMS & inventory records

Record experiments and samples correctly in lab systems.

20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.

Typical applications
Benchling · LabArchives · eLabFTW · SENAITE
Class GUI-need label
G2 · GUI-essential

Task proposals · 20

Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.

R08.12.01.01-T01 Create a qPCR experiment from template in eLabFTW with linked items

Original instruction

In the sandbox eLabFTW instance (test account in input/creds.txt), create an experiment from template 'qPCR run' titled 'qPCR GAPDH knockdown 2026-10-14', fill the Steps from input/run_sheet.pdf, link inventory items 'Primer GAPDH-F' and 'Primer GAPDH-R', attach input/plate_layout.xlsx, tag it 'project-K7' and set status 'Success'.

Platforms
web · Any OS (web)
Applications
eLabFTW
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms drag · Drag & drop
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

Checker queries the eLabFTW DB read-only: experiment exists with exact title, template origin, 6 steps text, 2 item links, attachment hash, tag and status. Agent account has no API key; API access logs must be empty for the session.

Proposed partial credit

15% created from template; 25% steps; 25% item links; 35% attachment, tag and status (3:4)

R08.12.01.01-T02 Register six water samples from a chain-of-custody form in SENAITE

Original instruction

In sandbox SENAITE LIMS, register the 6 samples on input/coc_2026-118.pdf for client 'Riverside Utility': sample type 'Drinking water', sampling date/time, sampling point, container and the analysis profile 'Metals-ICP' (or the individual analyses ticked on the form). Then receive all 6 samples.

Platforms
web · Any OS (web)
Applications
SENAITE
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
type · Text & forms click · Point & click vis · Visual judgment
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Checker reads SENAITE catalog via a read-only admin session: 6 samples for the client with exact field values and analysis sets per the form (2 forms tick extra analyses), all in state 'sample_received'. JSON API use by the agent account is blocked and logged.

Proposed partial credit

20% 6 samples created; 35% field values exact; 30% analysis sets; 15% received state

R08.12.01.01-T03 Enter ICP results for a 20-analysis worksheet with a retest in SENAITE

Original instruction

SENAITE worksheet WS-0042 holds 20 metal analyses. Enter the results from instrument printout input/icp_run_118.pdf (units mg/L, 3 values below LOQ entered as '<0.005'), submit them, then request a retest for the one Pb result that the printout flags 'QC fail' and enter the rerun value from page 2.

Platforms
web · Any OS (web)
Applications
SENAITE
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
type · Text & forms click · Point & click nav · Pan/zoom/scroll
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Analysis objects: 20 results equal the printout (detection-limit operand stored for <LOQ), states 'to_be_verified'; one retest exists for the flagged Pb with the rerun value and the original is 'retracted'. Instrument import via file upload is disabled in this sandbox. senaite.jsonapi is not installed and the agent lacks the Manager role (no ZMI).

Proposed partial credit

50% results correct (equal share each); 15% LOQ handling; 20% retest created and filled; 15% states

R08.12.01.01-T04 Register five plasmids from GenBank files in the Benchling registry

Original instruction

In the Benchling sandbox tenant, register the 5 plasmids in input/plasmids/*.gb into registry schema 'Plasmid' (folder 'Lab Stocks'), filling backbone, bacterial resistance, insert gene and promoter from the GenBank annotations, and link each to its parent strain entity named in input/parents.csv.

Platforms
web · Any OS (web)
Applications
Benchling
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms drag · Drag & drop
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Checker uses an admin read-only API token: 5 registered entities with schema 'Plasmid', sequences identical to the .gb files, field values exact, parent links correct. Agent user has no API token; audit log shows only UI actions.

Proposed partial credit

25% 5 entities registered with sequences; 40% schema fields; 20% parent links; 15% correct folder

R08.12.01.01-T05 Place 24 antibody aliquots into a freezer box per a plate map in Benchling

Original instruction

In Benchling Inventory create 24 tube containers holding 50 uL aliquots of entity 'AB-0042 anti-CD3' (lot from input/lot.txt) and place them in box 'FZ2-R3-B07' (9x9) following input/layout.xlsx, which skips column 5 and row E.

Platforms
web · Any OS (web)
Applications
Benchling
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Checker queries containers: 24 tubes with entity, volume and lot exact; each at the exact box position from the layout; no tubes in column 5 or row E; box has no other new contents. No Benchling API key or app credentials are issued to the test user and the VM has no shell, so containers can only be created in the browser; the tenant audit trail confirms each change came from the web session.

Proposed partial credit

25% 24 containers created; 20% contents and volume; 45% positions exact (proportional); 10% no stray placements

R08.12.01.01-T06 Create a dated LabArchives notebook page with files and request a witness

Original instruction

In the LabArchives sandbox notebook 'Smith Lab', create folder 'Project X/2026-Q4' and a page '2026-10-14 Western blot rerun' with a rich-text entry summarizing the 3 protocol deviations in input/deviations.txt, attach input/blot.tif and input/quant.xlsx, then sign the page and request witnessing by user 'pi_test'.

Platforms
web · Any OS (web)
Applications
LabArchives
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms drag · Drag & drop
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
2 / 5

Proposed verifier

Checker reads the notebook via an admin export: folder path and page title exact; text entry mentions all 3 deviations (keyword match); 2 attachments with matching hashes; page signed by the test user with a pending witness request to pi_test. No LabArchives API access key (akid) is issued and the VM has no shell, so the page can only be built in the browser; the page audit trail must show the test user's web-session entries.

Proposed partial credit

20% folder and page; 25% deviations text; 25% attachments; 30% signed and witness requested

R08.12.01.01-T07 Create an antibody resource category with custom fields and import 15 items

Original instruction

In eLabFTW, create a resources category 'Antibody' whose template has extra fields host (select: mouse/rabbit/goat), clone (text), dilution (text), lot (text) and expiry (date). Then add the 15 antibodies listed in input/antibodies.xlsx as items of that category with all fields filled.

Platforms
web · Any OS (web)
Applications
eLabFTW
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

DB check: category exists with 5 extra fields of the right types and select options; 15 items in the category with field values equal to the spreadsheet. CSV import is allowed only through the UI importer; API keys absent.

Proposed partial credit

30% category and field definitions; 10% 15 items exist; 60% field values (proportional)

R08.12.01.01-T08 Record plate-reader results in a structured Benchling notebook table

Original instruction

In Benchling create notebook entry 'ELISA IL-6 run 3' from template 'ELISA', fill its Results table (schema 'ELISA Result') with OD450 values for the 24 samples from input/platereader_export.xlsx, link each row to its registered sample entity, and submit the entry for review to 'reviewer_test'.

Platforms
web · Any OS (web)
Applications
Benchling
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
type · Text & forms click · Point & click multi · Cross-app
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Admin read-only query: entry from template; 24 result rows with OD values exact and sample links to the correct entities (by plate well mapping); entry state 'In review' with the right reviewer. No Benchling API key or app credentials are issued to the test user and the VM has no shell; the audit trail confirms the entry was filled in the web session.

Proposed partial credit

15% entry from template; 45% OD values; 25% sample links; 15% submitted for review

R08.12.01.01-T09 Configure a calculated nitrate analysis service in SENAITE

Original instruction

In SENAITE setup create analysis service 'Nitrate (as N)' with keyword NO3N, unit mg/L, method 'Cd reduction', 3 decimal precision, and a calculation from interim field 'Abs' using slope/intercept from input/calibration.pdf. Add specification limits for 'Drinking water' (max 10 mg/L) and add the service to profile 'Nutrients'.

Platforms
web · Any OS (web)
Applications
SENAITE
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

Setup objects checked: service fields exact; calculation formula evaluates correctly for 5 test Abs values; specification range present for the sample type; profile includes the service. A test sample created by the checker returns the expected computed result. senaite.jsonapi is not installed, the test user lacks the Manager role (no ZMI or plone.restapi) and the VM has no shell, so setup is reachable only through the SENAITE web UI.

Proposed partial credit

25% service fields; 35% calculation correct; 20% specification; 20% profile

R08.12.01.01-T10 Correct a wrong concentration in an eLabFTW entry with an audit trail

Original instruction

Experiment 'Lysis buffer prep 2026-10-02' in eLabFTW lists Tris at 500 mM; the correct value is 50 mM. Edit the entry to fix the value, add a comment stating the reason, and lock the experiment. Do not delete or recreate the entry.

Platforms
web · Any OS (web)
Applications
eLabFTW
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
2 / 5

Proposed verifier

DB: same experiment ID with body now containing '50 mM' and not '500 mM'; a new revision exists with the previous body preserved; comment with a reason present; locked flag set by the test user. The test account has no eLabFTW API key and key creation is disabled by the admin; request logs must show the edit came from the browser session.

Proposed partial credit

40% value corrected in same entry; 20% revision history preserved; 20% reason comment; 20% locked

R08.12.01.01-T11 Fulfil three sample requests in Benchling with new aliquots

Original instruction

Three open Benchling Requests ask for aliquots of registered cell lines (input/requests.pdf summarizes them). For each, create the requested number of cryovial aliquots from the parent stock container, decrement the parent volume, place aliquots in the requester's box, attach them to the request and mark the request Completed.

Platforms
web · Any OS (web)
Applications
Benchling
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms drag · Drag & drop
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
4 / 5

Proposed verifier

Checker queries requests and inventory: each request Completed with the right number of linked aliquot containers; aliquot contents and volumes exact; parent volumes decremented by the total; positions in the requester boxes free of conflicts. No Benchling API key or app credentials are issued to the test user and the VM has no shell, so requests can only be fulfilled in the browser; the audit trail confirms web-session changes.

Proposed partial credit

30% aliquots created per request; 20% parent volumes; 20% placement; 30% requests linked and completed

R08.12.01.01-T12 Partition soil samples into preserved containers in SENAITE

Original instruction

Three received samples (S-26-0311..0313) need partitioning for metals (HNO3-preserved 250 mL), nutrients (H2SO4-preserved 250 mL) and VOCs (40 mL vial, no headspace). Create the partitions in SENAITE assigning each analysis to the right partition per input/partition_rules.pdf, and print container labels as PDF to output/labels.pdf.

Platforms
web · Any OS (web)
Applications
SENAITE
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click drag · Drag & drop vis · Visual judgment
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
4 / 5

Proposed verifier

9 partitions exist (3 per sample) with correct container type and preservation; every analysis is on exactly one partition per the rules; labels PDF contains 9 partition IDs (text extraction). senaite.jsonapi is not installed, the test user lacks the Manager role (no ZMI or plone.restapi) and the VM has no shell, so partitions can only be made in the SENAITE web UI.

Proposed partial credit

30% 9 partitions; 25% containers and preservation; 35% analysis assignment; 10% labels

R08.12.01.01-T13 Reconcile a physical inventory count against eLabFTW stock records

Original instruction

input/inventory_count_2026-10.xlsx lists a freezer audit of 60 reagents. Compare it with eLabFTW items in category 'Reagent': update quantity fields for the 9 discrepancies, set status 'Disposed' for the 2 items not found, add the 3 found items missing from eLabFTW, and write output/reconciliation.csv (item, old, new, action).

Platforms
web · Any OS (web) win · Windows mac · macOS
Applications
eLabFTW · Excel
Interfaces
brow · Browser desk · Desktop GUI
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
multi · Cross-app type · Text & forms click · Point & click
Work product
WP08 · App/system state
Conditions
multiapp · Multi-app long · Long-horizon
Proposed difficulty
5 / 5

Proposed verifier

DB check of all 60 reagent items: quantities equal the audit, 2 disposed statuses, 3 new items with correct fields; no other items changed; reconciliation.csv rows match the gold action list. The test account has no eLabFTW API key and key creation is disabled by the admin; the VM has no shell, and request logs must show only browser-session edits.

Proposed partial credit

35% quantity updates; 15% disposals; 20% new items; 30% no collateral edits and CSV (equal halves)

R08.12.01.01-T14 Book microscope time in the eLabFTW scheduler around existing bookings

Original instruction

Book resource 'Confocal LSM-900' in the eLabFTW scheduler for three 2-hour sessions next week (Mon-Fri, 09:00-18:00) without overlapping existing bookings, at least one day apart, and linked to experiment 'Live imaging HeLa-GFP'. Titles 'Live imaging 1'..'3'.

Platforms
web · Any OS (web)
Applications
eLabFTW
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop click · Point & click nav · Pan/zoom/scroll
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Scheduler table: 3 new events for the resource, each exactly 2 h within working hours next week, no overlap with existing events, on 3 distinct non-adjacent days, linked to the experiment, titles exact. The test account has no eLabFTW API key and key creation is disabled by the admin; the VM has no shell, and request logs must show the bookings came from the browser session.

Proposed partial credit

30% 3 events of 2 h; 30% no overlap and hours; 20% spacing; 20% links and titles

R08.12.01.01-T15 Fill a 96-well plate with registered samples and concentrations in Benchling

Original instruction

Create a 96-well plate 'DRC-2026-14' in Benchling Inventory and fill wells with registered compound entities and concentrations per input/plate_map.xlsx (8-point 3-fold dilution series, 12 compounds, controls in column 12), with 100 uL per well.

Platforms
web · Any OS (web)
Applications
Benchling
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
drag · Drag & drop click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app long · Long-horizon
Proposed difficulty
4 / 5

Proposed verifier

Admin query of the plate: 96 wells with entity, concentration (units uM) and volume exactly per the map; control wells correct; empty wells empty. No Benchling API key or app credentials are issued to the test user and the VM has no shell, so wells can only be filled in the browser; the audit trail confirms web-session changes.

Proposed partial credit

15% plate created; 45% entities per well; 30% concentrations; 10% volumes

R08.12.01.01-T16 Invalidate a published result and reissue the certificate in SENAITE

Original instruction

Sample S-26-0207 was published with a transcription error in Lead (0.12 instead of 0.012 mg/L). In SENAITE invalidate the sample with reason 'Transcription error', enter the corrected value on the retest sample, verify it with the second test account (input/creds.txt), and publish a new certificate. Stop before emailing the client.

Platforms
web · Any OS (web)
Applications
SENAITE
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms
Work product
WP08 · App/system state
Conditions
risk · Irreversible/risky
Proposed difficulty
5 / 5

Proposed verifier

Original sample state 'invalid' with the reason; retest sample exists with corrected Pb and other analyses copied; verified by the second user (not the submitter); new COA PDF version contains 0.012; no outbound email in the mail sink. senaite.jsonapi is not installed, the test users lack the Manager role (no ZMI or plone.restapi) and the VM has no shell, so the web UI is the only route.

Proposed partial credit

25% invalidated with reason; 25% retest correct; 20% verified by second user; 30% new COA with no email sent (2:1)

R08.12.01.01-T17 Link a chain of prior experiments and add a dated checklist in eLabFTW

Original instruction

Create eLabFTW experiment 'CRISPR KO line validation - summary' that links, in chronological order of their dates, the 4 experiments tagged 'KO-line-7', and add steps for the 5 remaining validation tasks from input/plan.txt with the deadlines given there. Category 'Summary'.

Platforms
web · Any OS (web)
Applications
eLabFTW
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms nav · Pan/zoom/scroll
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

DB: new experiment with category; experiment links to exactly the 4 tagged entries, ordering in body list chronological; 5 steps with exact text and deadline dates. The test account has no eLabFTW API key and key creation is disabled by the admin; the VM has no shell, and request logs must show only browser-session edits.

Proposed partial credit

20% entry and category; 40% correct links; 15% chronological order; 25% steps with deadlines

R08.12.01.01-T18 Control pair of T07: create the antibody category and items via the eLabFTW API

Original instruction

Control pair of T07. Using the eLabFTW REST API v2 with the API key in input/apikey.txt, create the same 'Antibody' resource category with 5 extra fields and the 15 items from input/antibodies.xlsx. No browser is available.

Platforms
lin · Linux mac · macOS
Applications
eLabFTW · Python
Interfaces
api · API/SDK term · Terminal
GUI need · task setting
AP · API-sufficient
GUI need · ordinary access
AP · API-sufficient
Interaction primitives
code · Code & commands
Work product
WP08 · App/system state
Conditions
None declared
Proposed difficulty
3 / 5

Proposed verifier

Identical DB checks to T07 (category fields, 15 items, values), enabling a direct GUI-vs-API comparison on the same fixture.

Proposed partial credit

30% category and fields; 10% 15 items; 60% values (proportional)

R08.12.01.01-T19 Record a cell-line thaw workflow run in Benchling

Original instruction

In Benchling Workflows, complete the open task 'Thaw HEK293-K7 bank vials' for 6 vials: record post-thaw viable cell density and viability from input/counter_printout.pdf, mark vial 4 as failed (viability below 70%), create output flask containers for the 5 passing vials, and complete the stage run.

Platforms
web · Any OS (web)
Applications
Benchling
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
type · Text & forms click · Point & click nav · Pan/zoom/scroll
Work product
WP08 · App/system state
Conditions
licensed · Licensed pro app
Proposed difficulty
3 / 5

Proposed verifier

Workflow task status completed; 6 result rows with exact counts and viabilities; vial 4 flagged failed; 5 output containers linked to the stage run with correct source vials. No Benchling API key or app credentials are issued to the test user and the VM has no shell, so the run can only be completed in the browser; the audit trail confirms web-session changes.

Proposed partial credit

40% results entered; 15% failure flagged; 30% outputs created and linked; 15% run completed

R08.12.01.01-T20 Take a soil batch from receipt to published certificates in SENAITE

Original instruction

A batch of 10 soil samples arrived (input/coc_soil.pdf). In SENAITE register and receive them with analyses pH, moisture and Cu; add a method blank and a duplicate to the worksheet; enter results from input/results_soil.xlsx; submit as analyst_test, verify as reviewer_test, and publish one certificate per sample. Stop before emailing.

Platforms
web · Any OS (web)
Applications
SENAITE
Interfaces
brow · Browser
GUI need · task setting
G2 · GUI-essential
GUI need · ordinary access
G2 · GUI-essential
Interaction primitives
click · Point & click type · Text & forms nav · Pan/zoom/scroll
Work product
WP08 · App/system state
Conditions
long · Long-horizon risk · Irreversible/risky
Proposed difficulty
5 / 5

Proposed verifier

10 samples published; worksheet contains 30 routine analyses plus blank and duplicate reference analyses; result values exact; submitter and verifier users differ per audit log; 10 COA PDFs; no email in the mail sink. senaite.jsonapi is not installed and the test users lack the Manager role (no ZMI or plone.restapi), so the web UI is the only route.

Proposed partial credit

15% registration; 15% worksheet with QC; 30% results; 40% submit/verify by right users and certificates without emailing (equal halves)

Mapped research references

No benchmark reference is assigned to this class in the current mapping.

In this workflow family