Agent skill

Delegated File Edit Verification

by HKUDS in HKUDS/OpenSpace

Verify delegated file-edit results directly in the target workspace before accepting success, and rerun with absolute paths if the delegate's report conflicts with reality.

MITAuto-check passedDocuments & Office

Install Delegated File Edit Verification

skills CLI
$ npx skills add HKUDS/OpenSpace --skill delegated-file-edit-verification -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install HKUDS/OpenSpace delegated-file-edit-verification --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/HKUDS/OpenSpace.git skills-src && mkdir -p .claude/skills && cp -r skills-src/benchmarks/gdpval/skills/delegated-file-edit-verification .claude/skills/delegated-file-edit-verification && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
delegated-file-edit-verification
GitHub stars
7.8k
Token cost
~2.2k tokens
SKILL.md length
1,066 words
Files
2
Skills in repo
199
Repo updated
First seen
Licence
MIT

At a glance

Verify delegated file-edit results directly in the target workspace before accepting success, and rerun with absolute paths if the delegate's report conflicts with reality.

  • Works in 7 steps: Record the intended output contract… → Delegate with explicit acceptance criteria → Verify immediately after the delegate… → …
  • Tasks that involve Excel spreadsheets
  • SKILL.md covers Core principle, When to use this, Procedure and Verification checklist, plus 9 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Delegated File Edit Verification is an agent skill from HKUDS/OpenSpace. Verify delegated file-edit results directly in the target workspace before accepting success, and rerun with absolute paths if the delegate's report conflicts with reality.

Its SKILL.md is about 2.2k tokens, which your agent loads only when the skill is triggered. The skill folder holds 1 other file.

It sits in Documents & Office, covering Excel spreadsheets. The repository describes itself as: "OpenSpace: The Skill Management Layer for AI Agents" -- https://open-space.cloud/. The licence is MIT.

When your agent uses it

  • Tasks that involve Excel spreadsheets

Example prompts

  • “/delegated-file-edit-verification”

Requirements

  • Python 3

Workflow steps

7 steps, taken from the step headings in SKILL.md.

  1. Record the intended output contract before delegating
  2. Delegate with explicit acceptance criteria
  3. Verify immediately after the delegate reports success
  4. Prefer direct machine checks over narrative summaries
  5. If verification conflicts with the delegated summary, mark the task incomplete
  6. Re-run with absolute paths and constrained scope
  7. Re-verify after the rerun

What it can do on your machine

Read from SKILL.md and the folder at commit 3827781. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Delegated File Edit Verification loads about 2.2k tokens when it runs. Until then it costs about 51 tokens; SKILL.md has 1,066 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~51
When it runs · the whole SKILL.md, loaded when a task matches
~2.2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from HKUDS/OpenSpace at commit 3827781, republished under its MIT licence (© HKUDS). 1,066 words, ~2,194 tokens.

Download SKILL.mdSave it as .claude/skills/delegated-file-edit-verification/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
delegated-file-edit-verification
description
Verify delegated file-edit results directly in the target workspace before accepting success, and rerun with absolute paths if the delegate's report conflicts with reality.

Delegated File Edit Verification

Use this workflow whenever another agent, subprocess, or delegated tool claims it successfully created or modified a file. The goal is to prevent false acceptance of incomplete, misplaced, or incorrect edits.

This skill is especially useful for structured files such as spreadsheets, CSVs, databases, and generated artifacts, but the same pattern applies to any delegated file-edit task.

Core principle

Never accept a delegated "done" message on summary alone.

Immediately verify the result yourself in the target workspace using direct commands that inspect:

  • file existence
  • exact path
  • expected structure
  • expected size or counts
  • key headers, fields, or identifiers

If the delegate's summary and your verification disagree, treat the task as incomplete. Re-run the task using absolute paths and explicit acceptance checks.

When to use this

Use this workflow when:

  • a sub-agent says it wrote or updated a file
  • the task depends on a specific workspace or output location
  • the output is structured and can be checked mechanically
  • prior executions showed path confusion, stale files, or inconsistent summaries
  • correctness matters more than trusting the delegate's narrative

Procedure

1. Record the intended output contract before delegating

Before handing off the task, write down the minimum facts that must be true if the task succeeds.

Include:

  • absolute target path
  • required file name
  • required file type
  • expected sheets, tabs, tables, or sections
  • expected row counts or approximate size
  • required headers or columns
  • any sentinel values that should appear

Example contract:

  • file exists at /workspace/results/report.xlsx
  • workbook contains sheets summary and data
  • data sheet has more than 1,000 rows
  • headers include id, created_at, status

Do not rely on vague goals like "updated spreadsheet correctly."

2. Delegate with explicit acceptance criteria

When delegating, include:

  • the absolute input path
  • the absolute output path
  • the exact required structure
  • a request to report what was changed
  • a request to include paths actually used

Good delegation prompt:

  • Read /workspace/input/source.xlsx
  • Write the cleaned workbook to /workspace/output/cleaned.xlsx
  • Ensure sheets summary and records exist
  • Ensure records includes headers id, date, amount
  • Report the exact absolute path written and row counts per sheet
3. Verify immediately after the delegate reports success

Do not move on. Run direct checks in the target workspace right away.

For all files, verify:

  • the file exists at the expected absolute path
  • modification time is recent
  • file size is plausible
  • there is not a second similarly named file in another directory

For structured files, verify the internal structure too:

  • workbook sheet names
  • CSV header row
  • row counts
  • required columns
  • sample records or sentinel values
4. Prefer direct machine checks over narrative summaries

Trust commands and file inspection, not the delegate's prose.

Examples of useful checks:

  • list the exact file path
  • print sheet names
  • count rows
  • print header names
  • inspect a few sample rows
  • compare timestamps before and after
5. If verification conflicts with the delegated summary, mark the task incomplete

Examples of conflicts:

  • delegate says file exists, but it does not
  • delegate claims sheet data exists, but workbook contains different sheets
  • reported row count differs materially from actual count
  • expected headers are missing
  • output was written to a relative or unexpected location
  • the delegate inspected one file but wrote another

When this happens:

  • do not accept the result
  • state the exact mismatch
  • rerun the task
  • require absolute paths for every input and output
  • require a final report tied to those exact paths
6. Re-run with absolute paths and constrained scope

The most common recovery is to eliminate ambiguity.

On rerun:

  • provide absolute paths only
  • restate the acceptance contract
  • instruct the delegate to verify its own output before reporting success
  • ask it to print the exact path and structural facts used

Example rerun instruction:

  • Re-run using only these absolute paths:
    • input: /workspace/input/source.xlsx
    • output: /workspace/output/cleaned.xlsx
  • Before reporting success, verify:
    • file exists
    • workbook sheets are summary and records
    • records row count exceeds 1000
    • headers include id, date, amount
  • Report the exact output path, sheet names, row counts, and headers
Show full SKILL.md (418 more words)Show less
7. Re-verify after the rerun

Repeat your independent verification. Only accept the task when your checks match the expected contract.

Verification checklist

Use this checklist before accepting a delegated file-edit result:

  • I know the exact absolute output path.
  • I confirmed the file exists there.
  • I checked modification time or other evidence it was freshly written.
  • I verified the file's internal structure.
  • I checked row counts, record counts, or comparable size indicators.
  • I checked key headers, fields, or identifiers.
  • My direct inspection matches the delegate's report.
  • If there was any mismatch, I reran with absolute paths.

Examples

Example: generic file existence check

Use direct shell checks first.

text
ls -l /workspace/output/result.xlsx
stat /workspace/output/result.xlsx
find /workspace -name 'result.xlsx' -o -name '*result*'

Interpretation:

  • confirm the file exists where expected
  • compare timestamps and sizes
  • detect misplaced outputs

Example: CSV verification

text
test -f /workspace/output/data.csv
head -n 3 /workspace/output/data.csv
python - <<'PY'
import csv
path = "/workspace/output/data.csv"
with open(path, newline="") as f:
    reader = csv.reader(f)
    rows = list(reader)
print("rows_including_header:", len(rows))
print("headers:", rows[0])
PY

Check:

  • file exists
  • headers are correct
  • row count is plausible

Example: spreadsheet verification

text
python - <<'PY'
from openpyxl import load_workbook
path = "/workspace/output/report.xlsx"
wb = load_workbook(path, read_only=True, data_only=True)
print("sheets:", wb.sheetnames)
for ws in wb.worksheets:
    print(ws.title, ws.max_row, ws.max_column)
    headers = [c for c in next(ws.iter_rows(min_row=1, max_row=1, values_only=True))]
    print("headers:", headers)
PY

Check:

  • expected sheets exist
  • row counts are plausible
  • key headers are present

Example: delegate summary conflict

Delegate says:

  • wrote /workspace/output/report.xlsx
  • created sheet data
  • row count is 2,500

Your verification shows:

  • file exists at ./report.xlsx instead
  • workbook has sheets Sheet1 and summary
  • no data sheet exists

Correct response:

  • mark the task incomplete
  • do not accept the delegated summary
  • rerun using absolute paths and explicit checks

Use a concise recovery message like this:

  • Your report conflicts with direct verification.
  • Expected output: /absolute/path/to/output.ext
  • Observed issue: file missing / wrong sheets / wrong row count / missing headers
  • Treating task as incomplete.
  • Re-run using only absolute paths.
  • Before reporting success, include:
    • exact output path
    • structure summary
    • counts
    • key headers

Anti-patterns to avoid

Do not:

  • accept "done" without checking the file yourself
  • trust relative paths when workspace ambiguity is possible
  • rely on a delegate's sample output without checking the actual artifact
  • accept a file just because a similarly named file exists somewhere
  • assume a workbook or dataset matches the report without verifying sheets and counts
  • treat conflicting evidence as "probably fine"

Why this works

Delegated file-edit failures often come from:

  • writing to the wrong directory
  • inspecting a different file than the one modified
  • stale files from earlier runs
  • mistaken row counts or sheet names
  • overconfident summaries that skip verification

This workflow catches those failures early by making acceptance depend on direct evidence in the target workspace.

Success criterion

A delegated file-edit task is complete only when:

  1. the expected artifact exists at the exact intended path, and
  2. independent verification confirms its required structure and content, and
  3. that verification agrees with the delegated report.

If any of these fail, rerun with absolute paths.

© HKUDS, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file in benchmarks/gdpval/skills/delegated-file-edit-verification of HKUDS/OpenSpace.

  • SKILL.md
  • .skill_id

Open the folder on GitHubat commit 3827781

Compare with similar skills

Delegated File Edit Verification next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Delegated File Edit Verification compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Delegated File Edit Verification this skillHKUDS/OpenSpace7.8k—~2.2kAutomated safety check: PassMIT
MarkitdownImCa0/just-laws78114 repos~3.2kAutomated safety check: NotesMIT
Managing Google Workspacetaylorwilsdon/google_workspace_mcp3.3k—~2.9kAutomated safety check: PassMIT
Data Table Managern8n-io/n8n207k—~2.3kAutomated safety check: PassCustom licence
Docx4jplutext/docx4j2.4k—~2.5kAutomated safety check: PassNone
Instrument Data To Allotropeaws-samples/amazon-bedrock-agents-healthcare-lifesciences2742 repos~2.7kAutomated safety check: PassApache-2.0

Similar skills

  • Markitdown

    ImCa0/just-laws

    Convert files and office documents to Markdown. An agent skill from ImCa0/just-laws.

    781 GitHub starsUsed in 14 repos~3.2k tokens
    Documents & OfficeAuto-check: notes
  • Managing Google Workspace

    taylorwilsdon/google_workspace_mcp

    Manages Google Workspace operations across 12 services (Gmail, Drive, Calendar, Docs, Sheets, Slides, Forms, Tasks, Contacts, Chat, Apps Script, Custom Search).

    3.3k GitHub stars~2.9k tokensUpdated yesterday
    Documents & OfficeAuto-check passed
  • Official

    Load before calling data-tables or parse-file. An agent skill from n8n-io/n8n.

    207k GitHub stars~2.3k tokensUpdated today
    Documents & OfficeAuto-check passed
  • Docx4j

    plutext/docx4j

    A skill your agent uses when writing Java code that creates, reads or edits Word (.docx), PowerPoint (.pptx) or Excel (.xlsx) files with docx4j — including generating documents, editing existing…

    2.4k GitHub stars~2.5k tokensUpdated 2 days ago
    Documents & OfficeAuto-check passed
  • Instrument Data To Allotrope

    aws-samples/amazon-bedrock-agents-healthcare-lifesciences

    Official

    Convert laboratory instrument output files (PDF, CSV, Excel, TXT) to Allotrope Simple Model (ASM) JSON format or flattened 2D CSV.

    274 GitHub starsUsed in 2 repos~2.7k tokens
    Documents & OfficeAuto-check passed
  • Cyber Ppt

    crazyykhllc-bit/CyberPPT

    当用户需要把 DOCX、PDF、TXT、XLSX、研究报告、业务材料或原始数据转成高密度、可编辑、咨询风格 PPTX 时使用;也适用于需要 SCR 论证、视觉风格探索、详细图表和渲染质检的 PPT。

    1.8k GitHub stars~10k tokensUpdated 2 mo ago
    Documents & OfficeAuto-check passed

More from HKUDS/OpenSpace

All 199 skills in this repo
  • Walks through producing a master audio track plus stems in Python, from checking a reference file and timing sections by BPM to effects, a zip archive and final verification.

    7.8k GitHub stars~2.9k tokensUpdated 1 mo ago
    Auto-check passed
  • Handle cascading data retrieval tool failures by falling back to embedded knowledge generation

    7.8k GitHub stars~765 tokensUpdated 1 mo ago
    Auto-check passed
  • Gives an agent a workaround when its code-execution sandbox keeps failing: save the Python script to a file and run it through the shell instead.

    7.8k GitHub stars~588 tokensUpdated 1 mo ago
    Auto-check passed
  • A recovery routine for agents whose sandboxed code runner keeps failing: save the Python script to disk, then run it through the shell and read the output.

    7.8k GitHub stars~652 tokensUpdated 1 mo ago
    Auto-check passed
  • Fallback ladder for failed sandboxed code runs, plus the habit of fixing the working directory first so generated files land in the right place.

    7.8k GitHub stars~1.1k tokensUpdated 1 mo ago
    Auto-check passed
  • Fallback workflow for executing Python code when executecodesandbox fails repeatedly

    7.8k GitHub stars~1.1k tokensUpdated 1 mo ago
    Auto-check passed

Questions about Delegated File Edit Verification

What does Delegated File Edit Verification do?

Verify delegated file-edit results directly in the target workspace before accepting success, and rerun with absolute paths if the delegate's report conflicts with reality. Delegated File Edit Verification is an agent skill from HKUDS/OpenSpace. Verify delegated file-edit results directly in the target workspace before accepting success, and rerun with absolute paths if the delegate's report conflicts with reality.

When should I use Delegated File Edit Verification?

Delegated File Edit Verification fits situations like: tasks that involve Excel spreadsheets.

How do I install Delegated File Edit Verification in Claude Code?

Run `npx skills add HKUDS/OpenSpace --skill delegated-file-edit-verification -a claude-code`. Or copy the skill folder (benchmarks/gdpval/skills/delegated-file-edit-verification in HKUDS/OpenSpace) into .claude/skills/delegated-file-edit-verification in your project. Claude Code loads it when a task matches its description.

How do I install Delegated File Edit Verification in Codex?

Run `npx skills add HKUDS/OpenSpace --skill delegated-file-edit-verification -a codex`. Or copy the skill folder (benchmarks/gdpval/skills/delegated-file-edit-verification in HKUDS/OpenSpace) into .agents/skills/delegated-file-edit-verification in your project. Codex loads it when a task matches its description.

Can I use Delegated File Edit Verification in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add HKUDS/OpenSpace --skill delegated-file-edit-verification -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/delegated-file-edit-verification, .gemini/skills/delegated-file-edit-verification, .github/skills/delegated-file-edit-verification and .opencode/skills/delegated-file-edit-verification in your project.

What does Delegated File Edit Verification need to run?

SKILL.md names no scripts, command-line tools or credentials: Delegated File Edit Verification is instructions for the agent only. Our summary lists: Python 3.

Does Delegated File Edit Verification access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Delegated File Edit Verification safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Delegated File Edit Verification use?

Delegated File Edit Verification is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Delegated File Edit Verification use?

About 2.2k tokens (SKILL.md is roughly 8.8k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Delegated File Edit Verification?

Skills that share tags, products or a category with Delegated File Edit Verification: Markitdown (ImCa0/just-laws, 781 stars), Managing Google Workspace (taylorwilsdon/google_workspace_mcp, 3.3k stars), Data Table Manager (n8n-io/n8n, 207k stars) and Docx4j (plutext/docx4j, 2.4k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Delegated File Edit Verification?

HKUDS (a GitHub organization) maintains it in HKUDS/OpenSpace, which has 7,754 GitHub stars. The repository holds 199 skills in this directory. The repository was last updated on August 12, 2026.

Source: HKUDS/OpenSpace on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.