feat: add llmfill skill — AI Word form filling

[niuhuluzhihao]

What this adds

A new llmfill Agent Skill that auto-fills .docx forms, questionnaires, surveys, and templates using reference documents, a private knowledge base (RAG), web search, or pure LLM generation. Built for document-automation and batch form-completion workflows: DDQs, due diligence, compliance, HR, and application forms.

Why

Form-filling is a repetitive, document-heavy task that benefits from bundled resources — a stdlib-only Python CLI, reference docs, and worked examples — which is exactly the progressive-disclosure shape Agent Skills are designed for. The skill wraps the LLMFill service behind a minimal-permission CLI so a Copilot coding agent can fill forms end-to-end (upload -> poll -> download -> report cost) without leaving its session.

What's included

skills/llmfill/
├── SKILL.md                 # on-demand instructions + machine-readable permissions frontmatter
├── scripts/
│   ├── llmfill.py            # Python 3.10+ stdlib-only CLI entrypoint
│   ├── api_client.py         # HTTP client (multipart, polling, error parsing)
│   └── config.py             # config store (~/.llmfill/config.json, 600 perms)
├── references/
│   ├── api-documents.md      # document fill API
│   ├── api-knowledge.md      # knowledge base API
│   ├── api-account.md        # account / config / whoami
│   ├── curl-guide.md         # curl-only path when Python is unavailable
│   └── errors.md             # unified error code table
├── examples/
│   ├── fill-a-form.md        # end-to-end fill walkthrough (with polling)
│   └── build-kb.md           # knowledge-base create + ingest walkthrough
└── tests/
    └── test_llmfill.py        # script-layer tests (network monkeypatched)

All bundled assets are well under the 5 MB per-asset limit (largest is ~36 KB).

Validation

• npm run skill:validate — llmfill passes; all 419 skills valid
• npm run build (npm start) — docs/README.skills.md updated with the llmfill row
• Re-ran npm start to confirm idempotency (no further diff)
• Folder name llmfill matches name: in SKILL.md frontmatter
• description is within the 10–1024 char limit
• Branch cut from main (not staged)
• PR targets main

Usage notes

• The skill calls a single default endpoint (https://www.llmfill.com); custom endpoints require explicit user approval (fail-closed, persisted to approved_origins).
• API tokens (aif_-prefixed) are long-lived credentials configured by the user in ~/.llmfill/config.json (600 perms); the agent never touches the token in interactive flows.
• Documents are uploaded to the remote service for processing — the skill prompts for user confirmation before uploading sensitive material.

Test plan

• gh skills install github/awesome-copilot llmfill installs cleanly (after merge)
• python scripts/llmfill.py --help prints available subcommands
• python scripts/llmfill.py whoami (with a configured token) returns account/quotas
• pytest skills/llmfill/tests/ passes locally (network monkeypatched)

🤖🤖🤖

🤖🤖🤖

[github-actions]

🔒 PR Risk Scan Results

Scanned 12 changed file(s).

Severity	Count
🔴 High	0
🟠 Medium	0
ℹ️ Info	4

Severity	Rule	File	Line	Match
ℹ️	skill-script-touched	skills/llmfill/scripts/api_client.py	1	skills/llmfill/scripts/api_client.py
ℹ️	skill-script-touched	skills/llmfill/scripts/config.py	1	skills/llmfill/scripts/config.py
ℹ️	skill-script-touched	skills/llmfill/scripts/llmfill.py	1	skills/llmfill/scripts/llmfill.py
ℹ️	skill-script-touched	skills/llmfill/tests/test_llmfill.py	1	skills/llmfill/tests/test_llmfill.py

This is an automated soft-gate report. Findings indicate review targets and do not block merge by themselves.

[github-actions]

🔍 Vally Lint Results

✅ All checks passed

Scope	Checked
Skills	1
Agents	0
Total	1

Severity	Count
❌ Errors	0
⚠️ Warnings	0
ℹ️ Advisories	0

Summary

Level	Finding
ℹ️	✅ llmfill (2/2 checks passed)
ℹ️	✓ [spec-compliance] All 1 skill(s) are spec-compliant.
ℹ️	✓ spec-compliance: All spec checks passed.
ℹ️	✓ [valid-refs] All file references across 1 skill(s) are valid.
ℹ️	✓ valid-refs: All file references resolve to existing files within the skill directory.
ℹ️	1 skill(s) linted, 1 passed

Full linter output

### Linting skills/llmfill
✅ llmfill (2/2 checks passed)
    ✓ [spec-compliance] All 1 skill(s) are spec-compliant.
        ✓ spec-compliance: All spec checks passed.
    ✓ [valid-refs] All file references across 1 skill(s) are valid.
        ✓ valid-refs: All file references resolve to existing files within the skill directory.

1 skill(s) linted, 1 passed