Delegated File Edit Verification logo

Delegated File Edit Verification

OrganizationPopular
HKUDS
delegated-file-edit-verification

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.

Overview

PublisherHKUDS
RepositoryOpenSpace
Skill namedelegated-file-edit-verification
Stars
7.7K
Forks
918
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by HKUDS on GitHub. Read the source before you install it.

Installation

Install the Delegated File Edit Verification AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/HKUDS/OpenSpace.git /tmp/OpenSpace
mkdir -p .claude/skills
cp -r /tmp/OpenSpace/benchmarks/gdpval/skills/delegated-file-edit-verification .claude/skills/delegated-file-edit-verification
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Delegated File Edit Verification in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Delegated File Edit Verification on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Delegated File Edit Verification is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

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

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

Recommended response template when verification fails

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.

Frequently asked questions

What does the Delegated File Edit Verification AI skill 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.

Why use Delegated File Edit Verification on TypingMind?

Because you install it once and use it with any model. Delegated File Edit Verification is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Delegated File Edit Verification in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/HKUDS/OpenSpace/tree/main/benchmarks/gdpval/skills/delegated-file-edit-verification. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Delegated File Edit Verification?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Delegated File Edit Verification?

As many as you like. As long as a model supports skills, you can use Delegated File Edit Verification with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Delegated File Edit Verification AI skill free?

Yes. It is published on GitHub by HKUDS under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇