- Home
- Skills
- Prompt & Skill Engineering
- skill authoring prompts
Works with the AI Agents you already use
skill authoring prompts
A prompt library and linter for authoring, refining, and validating agent skills that actually trigger and execute.
$5
skill authoring prompts
Example session with this skill installed
I have a workflow for auditing AWS S3 buckets for public access. Turn this into a SKILL.md and review it for common failure modes.
- Read your context and instructions
- Compiled the authoring prompts
I have processed your S3 audit workflow through the authoring pipeline. I defined the trigger situations, converted the steps into an imperative method, and identified three critical failure modes where agents typically hallucinate security states.
SKILL.md Draft: S3 Public Access Audit
| Section | Content Summary |
|---|---|
| Description | Triggers on "audit s3", "check bucket permissions", or "is this bucket public". Focuses on IAM vs Resource Policy conflicts. |
| Method | 1. List buckets. 2. Get PublicAccessBlock. 3. Inspect Bucket Policy for 'Principal: *'. 4. Compare results against organization whitelist. |
| Failure Modes | Detects "Assume Role" bypasses. Counter-rule: check effective policy, not just explicit 'Allow'. |
| Files | scripts/s3-checker.py, policies/whitelist.json. |
---
name: s3-security-audit
description: Use when auditing AWS S3 buckets for public exposure or verifying bucket policy compliance. Triggers on "audit s3", "check bucket permissions", or "is this bucket public".
---
# S3 Security Audit
## Method
1. **Identify Target**: List all buckets in the current region and identify those without `PublicAccessBlock` enabled.
2. **Policy Inspection**: For each bucket, retrieve the Bucket Policy and check for statements containing `"Effect": "Allow"` and `"Principal": "*"`.
3. **ACL Validation**: Check Bucket ACLs for `AllUsers` or `AuthenticatedUsers` URI permissions.
4. **Report**: Generate a table of buckets flagged as "Public", "Internal-Only", or "Whitelisted".
## Failure Modes
- **Ignoring Block Public Access (BPA)**: The agent assumes a 'Public' policy means the bucket is open, ignoring account-level BPA settings. **Counter-rule**: Check `GetPublicAccessBlock` status before flagging a bucket as public.
- **Policy Complexity**: Missing 'Deny' statements that override 'Allow'. **Counter-rule**: You must evaluate the full policy JSON; do not summarize until the final step.
Next steps
- Run
python3 scripts/skill-lint.py .to verify the frontmatter and file paths. - Add a "Hard Case" example in
examples/scenarios.mdinvolving a bucket with a public policy but an account-level block. - Verify that
scripts/s3-checker.pyexists in your local environment.
Connects securely to your tools. The creator never sees your data.
About this skill
Writing a SKILL.md that actually triggers is harder than it looks. Most skills fail because their descriptions are too vague to activate or their methods are just feature lists rather than executable instructions. This skill provides a structured framework to move from a raw workflow to a production-ready agent skill that passes human and automated reviews.
What it does
- Drafts triggers by identifying the specific moments and phrases that should activate a skill.
- Outlines methods using imperative, ordered steps with clear completion criteria for every stage.
- Identifies failure modes to anticipate how an agent might misapply instructions and provides specific counter-rules.
- Generates hard cases for examples, moving beyond easy scenarios to test judgment and edge cases.
- Lints mechanics via a Python script that checks frontmatter, file paths, and trigger budget.
How it works
- Use
trigger-description.mdto define the scope and activation phrases before writing the body. - Run
method-outline.mdto convert expertise into a checkable sequence of operations. - Apply the
de-slop.mdprompt to strip away machine-like prose and hollow marketing claims. - Execute
scripts/skill-lint.pyto catch mechanical errors like broken file paths or missing failure modes.
Frameworks & tools
The skill includes a Python-based linter script and a library of Markdown prompt templates designed for high-reasoning LLMs like Claude 3.5 Sonnet or GPT-4o.
Why this beats prompting it yourself
Generic prompts often produce "wish lists" instead of checkable instructions. This skill enforces a "situation-first" architecture and uses a dedicated linter to catch the mechanical lies—like broken file references—that manual prompting misses.
Use cases
- Converting a repetitive manual terminal workflow into a shareable team skill.
- Fixing a "dead" skill that exists in the folder but never activates during sessions.
- Pre-flighting a skill submission to ensure it passes marketplace quality standards.
- Auditing a library of internal prompts to remove low-value "slop" and generic advice.
Known limitations
The linter checks mechanics only; it cannot verify if the logic of your method is technically sound. You must provide the initial domain expertise for the prompts to be effective.
Linter validates mechanics only, not logical correctness. Requires domain expertise as input. prompts are optimized for high-reasoning models.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago