Skip to content

Commit afad436

Browse files
imran-siddiqueCopilot
imran-siddique
and
Copilot
committed
docs: add SECURITY.md with responsible disclosure policy
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
1 parent 247bcd0 commit afad436

File tree

1 file changed

+56
-110
lines changed

1 file changed

+56
-110
lines changed

‎SECURITY.md‎

Lines changed: 56 additions & 110 deletions
Original file line numberDiff line numberDiff line change
@@ -1,143 +1,89 @@
11
# Security Policy
22

3+
## About Agent OS
4+
5+
**Agent OS** is part of the [Agent Governance Ecosystem](https://github.com/imran-siddique) — a suite of open-source projects for building, orchestrating, and governing autonomous AI agents in enterprise environments.
6+
37
## Supported Versions
48

59
| Version | Supported |
610
| ------- | ------------------ |
7-
| 0.x.x | :white_check_mark: |
11+
| Latest | :white_check_mark: |
12+
| < Latest | :x: |
813

9-
Once Agent OS reaches 1.0, we will maintain security updates for the latest minor version.
14+
Only the latest release of Agent OS receives security updates. Users are strongly encouraged to stay up to date.
1015

1116
## Reporting a Vulnerability
1217

13-
**Please do not report security vulnerabilities through public GitHub issues.**
14-
15-
If you discover a security vulnerability in Agent OS, please report it responsibly:
16-
17-
### 1. Email (Preferred)
18-
19-
Send an email to: **security@agent-os.dev**
20-
21-
Include:
22-
- Description of the vulnerability
23-
- Steps to reproduce
24-
- Potential impact
25-
- Any suggested fixes (optional)
26-
27-
### 2. GitHub Security Advisories
28-
29-
You can also report via [GitHub Security Advisories](https://github.com/imran-siddique/agent-os/security/advisories/new).
30-
31-
### 3. Encrypted Communication
32-
33-
For sensitive reports, you may encrypt your message using our PGP key:
34-
35-
```
36-
-----BEGIN PGP PUBLIC KEY BLOCK-----
37-
[PGP key to be added]
38-
-----END PGP PUBLIC KEY BLOCK-----
39-
```
40-
41-
Key fingerprint: `[To be added]`
42-
43-
## Response Timeline
44-
45-
| Stage | Timeline |
46-
|-------|----------|
47-
| Initial acknowledgment | Within 48 hours |
48-
| Initial assessment | Within 1 week |
49-
| Fix development | Varies by severity |
50-
| Public disclosure | Coordinated with reporter |
51-
52-
## Severity Classification
53-
54-
We use the following severity levels:
55-
56-
| Severity | Description | Response Time |
57-
|----------|-------------|---------------|
58-
| **Critical** | Remote code execution, complete bypass of safety policies | 24-48 hours |
59-
| **High** | Partial bypass of safety policies, data exposure | 1 week |
60-
| **Medium** | Denial of service, information leakage | 2 weeks |
61-
| **Low** | Minor issues, hardening opportunities | Next release |
62-
63-
## Security Model
64-
65-
### What Agent OS Protects Against
66-
67-
Agent OS provides **application-level policy enforcement**:
68-
69-
- ✅ Deterministic policy checks on agent actions
70-
- ✅ SQL injection prevention (via policy rules)
71-
- ✅ File system access restrictions (configurable paths)
72-
- ✅ Rate limiting and resource controls
73-
- ✅ Action logging and audit trails (Flight Recorder)
74-
75-
### What Agent OS Does NOT Protect Against
18+
If you discover a security vulnerability in Agent OS, please report it responsibly.
7619

77-
Agent OS is middleware, not a security sandbox:
20+
**Email:** [security@imransiddique.com](mailto:security@imransiddique.com)
7821

79-
- ❌ Memory corruption or process isolation (use containers)
80-
- ❌ Compromised LLM providers
81-
- ❌ Network-level attacks
82-
- ❌ Supply chain attacks on dependencies
22+
**Please include:**
8323

84-
**For production deployments, we recommend:**
85-
- Running agents in isolated containers
86-
- Using network policies to restrict agent communication
87-
- Monitoring the Flight Recorder for anomalies
88-
- Regular dependency audits
24+
- A description of the vulnerability
25+
- Steps to reproduce the issue
26+
- Affected version(s)
27+
- Any potential impact assessment
8928

90-
## Security Best Practices
29+
> **Do not open a public GitHub issue for security vulnerabilities.**
9130
92-
### For Users
31+
## What to Expect
9332

94-
1. **Keep Agent OS updated** - Always run the latest version
95-
2. **Use restrictive policies** - Start with minimal permissions, expand as needed
96-
3. **Enable Flight Recorder** - Log all agent actions for audit
97-
4. **Review policies regularly** - Ensure policies match current requirements
98-
5. **Run in containers** - Isolate agents for defense in depth
33+
| Step | Timeline |
34+
| --------------------------- | ------------ |
35+
| Acknowledgment of report | Within 48 hours |
36+
| Initial assessment | Within 5 business days |
37+
| Fix development and testing | Within 30 days (critical), 90 days (non-critical) |
38+
| CVE assignment | If applicable, coordinated with the reporter |
39+
| Public disclosure | After fix is released, per responsible disclosure timeline |
9940

100-
### For Contributors
41+
We will keep you informed throughout the process and credit reporters (unless anonymity is requested).
10142

102-
1. **Sign commits** - Use `git commit -s` (DCO) and GPG signing
103-
2. **Review dependencies** - Check for known vulnerabilities before adding
104-
3. **Write secure code** - Follow OWASP guidelines
105-
4. **Add tests** - Include security-relevant test cases
106-
5. **Document security implications** - Note any security considerations in PRs
43+
## Responsible Disclosure Timeline
10744

108-
## Dependency Management
45+
We follow a **90-day responsible disclosure policy**:
10946

110-
We monitor dependencies for known vulnerabilities using:
111-
- GitHub Dependabot
112-
- Regular security audits
47+
1. Reporter submits vulnerability via the email above.
48+
2. We acknowledge receipt within **48 hours**.
49+
3. We work to develop and release a fix within **90 days**.
50+
4. Once the fix is released, the vulnerability may be publicly disclosed.
51+
5. If we are unable to fix the issue within 90 days, we will coordinate with the reporter on an appropriate disclosure timeline.
11352

114-
## Disclosure Policy
53+
## Scope
11554

116-
We follow coordinated disclosure:
55+
### In Scope
11756

118-
1. Reporter notifies us privately
119-
2. We confirm and assess the vulnerability
120-
3. We develop and test a fix
121-
4. We coordinate disclosure timing with the reporter
122-
5. Fix is released with security advisory
123-
6. Credit is given to the reporter (unless they prefer anonymity)
57+
- Source code of Agent OS
58+
- Third-party dependencies used by Agent OS
59+
- Configuration files and deployment templates
60+
- CI/CD pipeline configurations
61+
- Documentation that could lead to insecure usage
12462

125-
## Security Advisories
63+
### Out of Scope
12664

127-
Published security advisories are available at:
128-
[github.com/imran-siddique/agent-os/security/advisories](https://github.com/imran-siddique/agent-os/security/advisories)
65+
- Social engineering attacks against maintainers or users
66+
- Denial of Service (DoS/DDoS) attacks
67+
- Attacks requiring physical access
68+
- Issues in third-party services not controlled by this project
69+
- Vulnerabilities already reported and being addressed
12970

130-
## Hall of Fame
71+
## Security Best Practices for Users
13172

132-
We recognize security researchers who responsibly disclose vulnerabilities:
73+
- **Keep dependencies updated:** Regularly run dependency audits and update to the latest versions.
74+
- **Use environment variables for secrets:** Never hardcode credentials, API keys, or tokens in configuration files.
75+
- **Enable access controls:** Follow the principle of least privilege when configuring agent permissions.
76+
- **Review configurations:** Audit your deployment configurations against the provided security guidelines.
77+
- **Monitor for advisories:** Watch this repository for security advisories and release notes.
78+
- **Use signed commits:** Enable GPG or SSH commit signing to ensure code integrity.
79+
- **Run in isolated environments:** Use containers or sandboxed environments for agent workloads.
13380

134-
*No submissions yet - be the first!*
81+
## Agent Governance Ecosystem
13582

136-
## Contact
83+
Agent OS is part of the broader **Agent Governance Ecosystem**, which provides a unified framework for secure, observable, and compliant AI agent operations. Security policies are coordinated across all ecosystem projects to ensure consistent protection.
13784

138-
- Security issues: security@agent-os.dev
139-
- General questions: [GitHub Discussions](https://github.com/imran-siddique/agent-os/discussions)
85+
For ecosystem-wide security concerns, please contact [security@imransiddique.com](mailto:security@imransiddique.com).
14086

14187
---
14288

143-
*Last updated: February 2026*
89+
Thank you for helping keep Agent OS and the Agent Governance Ecosystem safe for everyone.

0 commit comments

Comments
 (0)