- Home
- Skills
- Prompt & Skill Engineering
- Aegis Core Governance
Aegis Core Governance
Turn complex AI work into bounded contracts with risk classification, approval gates, untrusted-input handling, and evidence-backed handoff.
Free
Works with the AI Agents you already use
Aegis Core Governance
Example session with this skill installed
Inspect config/app.json. I want the timeout changed from 30 to 45 seconds. Show me the exact diff first, do not write until I explicitly approve this specific change, and do not inspect .env or any other file.
- Read your context and instructions
- Compiled the aegis core governance
Outcome
PARTLY DONE — preview prepared; no file was changed.
Task contract
- Objective: change only the timeout value from 30 to 45 seconds.
- Scope: config/app.json only; exclude .env, secrets, dependencies, and unrelated files.
- Acceptance: the diff has one value change, the JSON remains valid, and no out-of-scope file changes.
- Risk: reversible local change.
Proposed diff
config/app.json: timeout 30 -> 45
Approval gate
Approve this exact one-value change to config/app.json? Approval is single-use and expires if the target or effect changes. Silence is not approval.
Evidence
Observed: source inspected and preview created.
Changed: none.
Not done: no write, upload, network request, or secret read.
Connects securely to your tools. The creator never sees your data.
About this skill
Aegis Core Governance is an instruction-led skill for bounded, reviewable AI work. It helps an agent distinguish read-only inspection from a consequential change, define a compact task contract, classify risk, pause at a specific approval gate, treat instructions found in files and tool output as untrusted data, and produce an evidence-based handoff.
Included in this package:
- A focused SKILL.md workflow for scope, permissions, approval, evidence, and verification.
- An optional Python 3.9+ standard-library contract checker.
- A JSON contract schema and a worked approval/injection example.
Use it for dry runs, repository or configuration changes, code execution, generated artifacts, external actions, or sensitive-data handling. Simple read-only questions can use the lightweight path.
This is the Core skill package, not the Aegis Stable runtime. It does not provide a technical sandbox, kill switch, model router, MCP server, scheduler, or host-level enforcement. Those controls remain dependent on the agent host. No network connection or external account is required.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Prompt & Skill Engineering.
Advisory only: it cannot create or override a host sandbox, kill switch, permission system, or scheduler. It does not route models, run an MCP server, deploy to production, or require network access. The optional helper validates contract structure; it does not decide whether a task is wise or authorized.
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 buying a skill.
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
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 16 days ago
- Free with an Agensi account
Nothing to install. It runs in your AI agent.
- Optional
Python 3.9+
Python only runs contract_check.py.
Permissions
The skill reads and writes only within a workspace root and targets explicitly approved for the current task. It does not grant access, require an account, or authorize network transmission. Terminal/Shell is used only when the host runs the optional local checker or an approved command; Read Files and Write Files apply only to the approved scope.