Source
What does the pinned source say?
Trust center
Keep evidence, judgment, and uncertainty visible
SkillSignal is not a security certifier. Our job is to trace sources, separate evidence, publish methods and limits, and show users what to inspect next.
- Evidence types
- 5
- Security certification
- No
- Corrections
- Public
Evidence model
Five evidence types answer different questions
More evidence is not automatically better. It should be independent, methodical, and never presented as a different evidence type.
Computed
What do deterministic rules observe?
Tested
What happened in a documented environment?
Editorial
What decision and trade-off does an editor see?
Feedback
What did maintainers or users report?
What we do
Trace, label, explain, and correct
Canonical identity
Resolve owner, repository, path, aliases, and duplicate sources before creating a canonical entity.
Pinned provenance
Keep source path, commit, license, collection time, and owner context visible.
Explainable calculations
Publish the components behind quality and ranking signals.
Visible limits
State what static checks, tests, and editorial reviews cannot prove.
Corrections
Provide a public route to report factual, attribution, or source errors.
Lifecycle
Move stale, deprecated, redirected, or unsafe entities out of active discovery.
Claims we can make
- A specific source file existed at a recorded commit.
- Configured static rules detected visible patterns.
- A documented test produced a result within its scope.
- An editorial page was reviewed on a visible date.
Claims we do not make
- A Skill is universally safe or secure.
- Repository popularity proves an individual Skill is effective.
- One successful test proves all environments will work.
- Source ownership alone makes every instruction official or correct.