Is foscomputerservices/fosmvvm-fluent-datamodel-generator safe?
The fosmvvm-fluent-datamodel-generator skill is a well-structured Swift/Vapor code generation guide with no malicious content, prompt injection attempts, or executable components. Canary file accesses observed during monitoring are attributed to the audit harness pre- and post-install verification routines, not to the skill, and all canary integrity checks passed. The only notable concerns are unaudited sibling skill dependencies and a downstream SQL injection risk in generated migration templates that is contingent on attacker-controlled input reaching the code generation step.
Category Scores
Findings (4)
LOW Canary file read events during install window -12 ▶
Sensitive canary files (.env, .ssh/id_rsa, .aws/credentials, .npmrc, .docker/config.json, GCloud credentials) were opened during the monitoring window. The early batch (1771917343) is attributable to the audit harness pre-install baseline check. The post-install batch (1771917365) aligns with audit system post-verification. No skill content instructs or could trigger these reads.
INFO Template placeholder {baseDir} in documentation reference -5 ▶
SKILL.md contains 'OpenClaw reference'. This is a template variable used in the OpenClaw skill system to substitute the skill base directory at runtime. It is not an instruction to fetch a URL or read an arbitrary path; it resolves to a local reference within the installed skill tree.
INFO Sibling skill dependencies not audited -9 ▶
The skill documents a dependency on fosmvvm-fields-generator and references fosmvvm-viewmodel-generator. These skills are not included in this audit package. Their security profiles are unknown and should be evaluated independently before installation in a complete workflow.
INFO Skill contains raw SQL patterns with unsafeRaw interpolation -4 ▶
The reference.md template demonstrates SQLKit raw SQL with (unsafeRaw:) string interpolation for identifiers. This is a legitimate Vapor/SQLKit pattern for table-level operations, but if an agent were manipulated to substitute attacker-controlled values into the schema string before generation, it could produce SQL injection in migration code. Risk is downstream in user-generated code, not in the skill itself.