Only the latest minor of v4.x receives security updates. Older minors are deprecated immediately on the next minor release.
Worked example: v4.32.0 is the current latest minor (released 2026-05-07). Once v4.33.0 ships, v4.32.x will be deprecated for security purposes immediately. Adopters consuming via make update should pin to the major line (v4.x) rather than a specific minor; adopters who require longer-than-rolling support windows for a specific minor should fork or adopt the same backport pattern themselves.
Use the Report a vulnerability button on the Security tab of this repo.
Or navigate directly to https://github.com/davidmatousek/tachi/security/advisories/new to open a private advisory.
Please do not open a public GitHub Issue for security vulnerabilities — public Issues broadcast the vulnerability before a fix is available.
- Description of the vulnerability
- Steps to reproduce
- Affected components (agents, commands, schemas, templates, scripts)
- Potential impact
Maintainers: this channel relies on the GitHub Private Vulnerability Reporting toggle being enabled in repo Settings → Security.
- Acknowledgment within 5 business days.
- Assessment within 1 week of acknowledgment.
- Fix or mitigation timeline communicated after assessment.
- Credit in the fix commit and release notes by default; anonymity available on request.
In-scope for private vulnerability reports:
- tachi codebase paths:
.aod/scripts/bash/,.claude/agents/,.claude/commands/,.claude/skills/,stacks/ - Stack-pack scaffolds as shipped:
stacks/*/scaffold/— vulnerabilities in tachi's default scaffolded templates are in-scope. - Schema definitions:
contracts/,schemas/ - Template content that shapes report generation
- tachi-shipped configuration files that reference external services or credentials
The following are not handled through this private disclosure channel:
- Claude Code itself → report to Anthropic.
- Third-party MCP servers → report to their maintainers.
- Adopter personalization data (
.aod/personalization.env,brands/*/) → adopter's responsibility. - Adopter-modified scaffold output (post-
make scaffoldcustomizations) → adopter's responsibility. - Threat-model accuracy concerns (false positives, missed threats) → file as regular GitHub Issues, not security advisories.