Secure coding improvements, pr number 1

[TomasTomecek]

I will send more with followup PRs. There were even more changes here but decided to split/cut them because I didn't like Claude's proposals.

Claude's detailed overview:

2. Add credential redaction to MCP gateway logs (HIGH)

Files: mcp_server/gateway.py, mcp_server/tests/unit/test_gateway.py

Added _redact() function with regex patterns matching common credential formats:

• GitLab PATs (glpat-...)
• Anthropic API keys (sk-ant-...)
• Google API keys (AIzaSy...)
• OAuth2 tokens in URLs (oauth2:...@)
• Generic long tokens (20+ chars with keywords like token/key/password)

The log_tool_call decorator now redacts all tool arguments and error messages before logging.

Test coverage: 20 unit tests covering pattern matching, edge cases, and real-world scenarios.

3. Add LITELLM_DEBUG environment variable guard (HIGH)

Files: agents/utils.py, templates/beeai-agent.env

set_litellm_debug() now checks for LITELLM_DEBUG env var before enabling debug mode. If not set, logs a warning and returns without enabling token-leaking debug output.

Added comprehensive documentation explaining the security risk and usage.

5. Add SecurityContext to K8s deployments (LOW)

Files: openshift/deployment-*.yml (10 files)

Replaced empty securityContext: {} with:

securityContext:
  runAsNonRoot: true
  seccompProfile:
    type: RuntimeDefault

Matches the pattern already used in cronjob manifests.

[gemini-code-assist]

Code Review

This pull request significantly enhances application security by implementing credential redaction in logs and strengthening Kubernetes deployment configurations. Specifically, a new _redact utility and associated regex patterns were introduced in mcp_server/gateway.py to prevent sensitive data like API keys and tokens from appearing in log outputs, with comprehensive unit tests added for validation. The set_litellm_debug function was updated to prevent token leakage by requiring an explicit environment variable for debug logging and adding a clear warning. Furthermore, all Kubernetes deployment files were updated to enforce runAsNonRoot: true and apply a default seccompProfile for improved runtime security. A review comment suggested making the _REDACT_PATTERNS a frozenset to ensure its immutability.

[TomasTomecek]

/gemini review

[gemini-code-assist]

Code Review

This pull request introduces several security enhancements, including the integration of detect-secrets into the pre-commit workflow, the addition of a redaction mechanism to mask sensitive credentials in logs, and the hardening of OpenShift deployments with non-root security contexts. Feedback focuses on refining the redaction regex to prevent destructive string replacement, consolidating duplicate sanitization logic into a shared utility, and strengthening unit tests to verify that redacted strings preserve their original structure.

[nforro]

I'm curious if this won't cause trouble when logging e.g. Celery task IDs (if we decide to use Celery).

[TomasTomecek]

great point Nikola; I expect that we'll update the regexes in the future or even the logging mechanism