Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only Temporal testing skill, but it gives under-scoped guidance for exporting production workflow histories into local and CI environments.
Review this skill carefully before using its replay-testing and CI examples. Prefer sanitized or synthetic histories, require explicit approval before exporting production histories, restrict credentials, prevent history files from entering logs/caches/artifacts, pin dependencies, bind local services to localhost, and treat docker-compose down -v as destructive.
resources/replay-testing.md:181Unsafe Transfer and Storage of Production Workflow Histories in CI
resources/local-setup.md:15Mutable and Insufficiently Pinned Third-Party Dependencies
resources/local-setup.md:11Development Services Exposed on All Host Interfaces with Default Database Credentials
The document includes docker-compose down -v as a reset step, which deletes Docker volumes and can erase local PostgreSQL/Temporal state. In a local setup guide this is not inherently malicious, but presenting the command without an explicit warning increases the chance of accidental data loss, especially if users adapt the compose file for non-ephemeral development data.
The clean Makefile target combines docker-compose down -v with removal of .pytest_cache, htmlcov, and .coverage, causing irreversible deletion of local test artifacts and persisted container data. Because make clean is a common, easy-to-run command, users may trigger broader deletion than expected if the target is not clearly documented as destructive.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
class BadWorkflow:
@workflow.run
async def run(self) -> dict:
response = requests.get("https://api.example.com/data") # External call!
return response.json()
# ✅ Deterministic
The guidance explicitly exports and stores production workflow histories to local files without warning that Temporal histories may contain workflow inputs, results, arguments, and other sensitive business data. In a testing skill, this can normalize unsafe handling of production data and lead to accidental retention, developer workstation exposure, or onward sharing in source trees, backups, or CI systems.
The CI example automates downloading production workflow histories onto shared CI runners and then uploads failure artifacts, but provides no warning about exposing production data in CI logs, workspaces, caches, or artifacts. In this context, replay data is more dangerous because CI environments often have broader access surfaces and longer-lived artifact retention than local development.
The example trigger "How do I mock activities?" is a generic natural-language request rather than a narrowly scoped invocation cue. In isolation, it could overlap with everyday conversation and does not include explicit constraints or exclusion conditions for when this skill should be selected.
No suspicious patterns detected.