chat-attachment-is-not-a-file
Skill1 starsChanged 32 days agoMITView on GitHub
Use when a task must read, edit, upscale, composite or train on an image or document the user pasted into chat.
In this repository121 other files it gives its agents.
- analysis-documentationSkill
- browser-automationSkill
- business-metrics-calculatorSkill
- claude-code-masterySkill
- codebase-designSkill
- codex-cli-specialistSkill
- cohort-analysisSkill
- cut-the-curveSkill
- data-narrative-builderSkill
- data-quality-auditSkill
- domain-modelingSkill
- focused-fixSkill
- funnel-analysisSkill
- gate-must-derive-its-criterionSkill
- grill-with-docsSkill
- gsd-add-testsSkill
- gsd-ai-integration-phaseSkill
- gsd-audit-fixSkill
- gsd-audit-milestoneSkill
- gsd-audit-uatSkill
- gsd-autonomousSkill
- gsd-captureSkill
- gsd-cleanupSkill
- gsd-code-reviewSkill
- gsd-complete-milestoneSkill
- gsd-configSkill
- gsd-debugSkill
- gsd-discuss-phaseSkill
- gsd-docs-updateSkill
- gsd-eval-reviewSkill
- gsd-execute-phaseSkill
- gsd-exploreSkill
- gsd-extract-learningsSkill
- gsd-fastSkill
- gsd-forensicsSkill
- gsd-graphifySkill
- gsd-healthSkill
- gsd-helpSkill
- gsd-importSkill
- gsd-inboxSkill
- gsd-ingest-docsSkill
- gsd-managerSkill
- gsd-map-codebaseSkill
- gsd-milestone-summarySkill
- gsd-mvp-phaseSkill
- gsd-new-milestoneSkill
- gsd-new-projectSkill
- gsd-ns-contextSkill
- gsd-ns-ideateSkill
- gsd-ns-manageSkill
- gsd-ns-projectSkill
- gsd-ns-reviewSkill
- gsd-ns-workflowSkill
- gsd-pause-workSkill
- gsd-phaseSkill
- gsd-plan-phaseSkill
- gsd-plan-review-convergenceSkill
- gsd-pr-branchSkill
- gsd-profile-userSkill
- gsd-progressSkill
- The first 60 shown.
The file
Files
A chat attachment has no path — and its filename lies
Extracted: 2026-08-03 Context: img2img, upscaling, compositing, LoRA datasets, PDF/table parsing — anything taking a file path.
Problem
The model can see a pasted image and describe it perfectly, which creates the illusion it can also process it. It cannot: gradio_client, PIL, and every CLI need a real path. Re-attaching the image does not help. This silently costs many round-trips if you promise work you can't start.
Solution
1. Search before asking. The user may already have saved it, under a name you'd never guess. Scan the likely folders filtered by recent mtime:
$cut = (Get-Date).AddHours(-2)
Get-ChildItem "$env:USERPROFILE\Downloads" -File -Include *.png,*.jpg,*.jpeg,*.webp |
Where-Object { $_.LastWriteTime -gt $cut } | Sort-Object LastWriteTimeWiden to Desktop / Pictures / OneDrive / the project folder before concluding it's missing. Note a name filter can miss it entirely — list the newest files regardless of name.
2. Ask once, concretely. Give an exact folder plus suggested filenames, and explicitly accept "save it anywhere and tell me the path". Point out they already know how if they've downloaded your outputs before.
3. Verify identity by INSPECTION, never by name or size.
- Bytes are not resolution. A 58 KB 512x512 crop outweighs a 13 KB 1024x531 full image — a "largest file wins" heuristic picks the wrong source. Sort by
Image.open(f).size, notLength. - The user's numbering is not yours. A file called
notext-1 (2).pngturned out to be image five. Map files to images by dimensions/aspect ratio. - Echo back what you matched (
name -> WxH -> which image) so a mismatch surfaces before you spend GPU on the wrong input.
When to Use
Whenever the deliverable requires consuming a user-supplied file and the only copy you've seen is in the conversation; or before selecting among several similarly named saved files.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

