Menu
Use when asked to create, generate, or scaffold a new Agent Skill. Don't use for modifying existing skills or writing general documentation.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when converting files to Markdown format for LLM consumption, document analysis, or text extraction. Supports PDF, Office documents, images, audio, HTML, and more via Microsoft's markitdown. Don't use for simple text file reads, existing Markdown files, or when the file is in a usable format.
Use when creating or modifying custom opencode commands in the `commands/` directory. Covers markdown command files, frontmatter options (description, agent, model, subtask), prompt templates with $ARGUMENTS, shell output injection, and file references.
Use when rendering mermaid diagrams to PNG, SVG, or PDF images. Handles both .mmd files and inline mermaid code blocks in markdown files. Don't use for creating mermaid syntax or GraphViz/DOT diagrams.
Use when creating a comprehensive architecture document for a codebase. First explores the codebase using the exploring-code skill, then writes a detailed architecture document. Don't use for quick documentation lookups or single-file questions: only use when asked to create or update architectural documents.
Use to do a thorough exploration of an existing codebase. Always use this when asked to explore a codebase/repo/project. Always use it when starting a task that requires deep knowledge of the codebase. Don't use for quick lookups or single-file questions.
You MUST use when creating, modifying, or reviewing GitHub Actions workflow files. For security reasons it is VITAL that you use this skill when working with third party actions (including `actions/checkout` and other official actions).
| name | create-skill |
| description | Use when asked to create, generate, or scaffold a new Agent Skill. Don't use for modifying existing skills or writing general documentation. |
| license | MIT |
| metadata | {"author":"Robert Hafner","source":"https://github.com/tedivm/opencode-config"} |
processing-pdfs, analyzing-data, testing-code^[a-z0-9]+(-[a-z0-9]+)*$ (1-64 chars, lowercase, hyphens as separators only, no leading/trailing/consecutive hyphens)helper, utils, tools) and reserved words (anthropic, claude)Always place skills in ./.agents/skills/<name>/:
mkdir -p ./.agents/skills/<name>/
Required fields:
name — must match directory namedescription — 1-1024 chars, written in third person, specific enough for correct activationOptional fields:
license — license name or reference to a bundled license filecompatibility — 1-500 chars, environment requirementsmetadata — arbitrary string-to-string key-value map (use author for the author name)Always add a source field under metadata pointing to the repository URL. Resolve it dynamically — do not hardcode:
git remote get-url origin to get the remote URLssh format (git@github.com:user/repo.git) to https (https://github.com/user/repo).git suffix if presentmetadata:\n source: <url> to the frontmatterDetermine author and license dynamically:
whoami to check the current usernametedivm:
license: MITmetadata.author: Robert HafnerGood: "Use when extracting text from PDFs, filling forms, or merging documents." Bad: "Extracts text and tables from PDF files, fills forms, and merges documents. Use when working with PDF files or when the user mentions PDFs, forms, or document extraction."
references/ files.Match specificity to the task's fragility:
Don't offer too many options. Provide a default with an escape hatch:
Use pdfplumber for text extraction.
For scanned PDFs requiring OCR, use pdf2image with pytesseract instead.
## Quick start
[Brief overview with the most common use case]
## Workflow
[Step-by-step numbered instructions]
## Advanced features
[Links to reference files for less common features]
Use one of these patterns to organize content:
Pattern 1: High-level guide with references
## Quick start
[Core instructions inline]
## Advanced features
**Form filling**: See [FORMS.md](FORMS.md)
**API reference**: See [REFERENCE.md](REFERENCE.md)
Pattern 2: Domain-specific organization
## Available domains
**Finance**: Revenue, billing → See [reference/finance.md](reference/finance.md)
**Sales**: Pipeline, accounts → See [reference/sales.md](reference/sales.md)
Pattern 3: Conditional workflow
## Document modification
**Creating new content?** → Follow "Creation workflow" below
**Editing existing content?** → Follow "Editing workflow" below
For complex multi-step tasks, provide a checklist the agent can track:
Copy this checklist and check off items as you complete them:
Task Progress:
For quality-critical tasks, implement validate-fix-repeat cycles:
1. Make your edits
2. Validate: python scripts/validate.py input/
3. If validation fails, fix and re-run
4. Only proceed when validation passes
5. Rebuild: python scripts/build.py input/ output/
For output-dependent skills, provide input/output pairs:
## Commit message format
**Example 1:**
Input: Added user authentication with JWT tokens
Output: feat(auth): implement JWT-based authentication
Follow this style: type(scope): brief description.
For strict output requirements, provide exact templates:
## Report structure
ALWAYS use this exact template:
# [Title]
## Executive summary
[One paragraph]
## Key findings
- Finding with data
For flexible guidance, provide a default and allow adaptation.
skill-name/
├── SKILL.md # Required: metadata + core instructions (<500 lines)
├── scripts/ # Executable code as tiny CLIs
├── references/ # Supplementary context (schemas, cheatsheets)
└── assets/ # Templates or static files
reference/guide.mdform_validation_rules.md, not doc2.mdMany skills are wrappers around existing CLI tools (gh, rsync, jq, git, etc.). Before writing instructions, research the CLI thoroughly:
<tool> --help, <tool> <subcommand> --help) to discover available flags, subcommands, and usage patternsDon't bundle or reimplement these CLIs — assume they're available and provide installation instructions in a separate file.
references/installation_guide.mdbrew install jq, pip install pdfplumber, etc.)analyze_form.py" (execute) vs "See analyze_form.py" (read as reference)README.md, CHANGELOG.md, or other human-facing documentationTest the description in isolation with an LLM:
Simulate execution: feed the full SKILL.md to an LLM and ask it to walk through a representative task step by step, flagging any ambiguities or missing steps.
Ask an LLM to attack the skill's logic: what happens on failure, unsupported configurations, missing dependencies?
SKILL.md is spelled in all capsname and descriptionname matches directory namemetadata.source contains the resolved repository HTTPS URLlicense is set (MIT for tedivm)metadata.author is set (Robert Hafner for tedivm)