Inspected source establishes explicit agent instruction mutation and guarded command rewriting, which warrant a capability warning under the supplied policy. Scanner labels do not establish credential theft or an install-time agent takeover, so publication blocking is unsupported.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review; previous stored version diff introduced dangerous source
Trigger
Running cleo provider inject, or invoking the configured heavy-command agent hook.
Impact
These capabilities can change agent instructions and command routing, including global instruction scope when explicitly requested.
Mechanism
The CLI forwards caller-selected instruction content and scope to core; the hook emits modified tool input subject to permission-mode and rule checks.
Evidence
dist/cli/index.jsdist/cli/hook-entry.jsAGENTS.md