sf-ai-agentforce-persona

Fail

Audited by Runlayer on Mar 13, 2026

Risk Level: HIGH
Scan Summary
Max Score
95%
Files
8
Flagged
4
Chunks
17
Flagged Files (4)
sf-ai-agentforce-persona/SKILL.mdHIGH
94.9%

Malicious tool definition detected

Constraint notes in the framework explain how earlier choices pull later ones. ## GENERATION GUARDRAILS These behavioral constraints apply when executing the persona design workflow: - **Context fields are agent design, not persona** — Name, role, audience, agent type, surface, use cases originate in agent design. Persona design imports or minimally gathers them.

Tool: sf-ai-agentforce-persona/SKILL.md [4/4] Description: and persona boundary scenarios. | **Scoring rules:** - Score each category independently. Provide a number and 1-2 sentences of justification. - Flag any inconsistencies between dimensions or settings (e.g., "Personable voice but Clinical Analyst tone — these may conflict in practice").

sf-ai-agentforce-persona/references/persona-encoding-guide.mdMEDIUM
83.6%

Tool passed security scan

Malicious tool definition detected

Tool: sf-ai-agentforce-persona/references/persona-encoding-guide.md [3/4] Description: by a dropdown - **Identity** — personality traits are far richer than 3 options **Recommendation:** Set the dropdown to whichever option is closest to your persona's Register + Voice combination. Then do the real persona work in Topic Instructions and Action Output Response Instructions. The dropdown is a global nudge; the instructions are where persona actually lives.

- **Instruction precedence:** The working assumption is that topic instructions supersede system instructions, but more testing is needed to evaluate this. If confirmed, the system prompt carries the baseline persona and topic instructions carry overrides — the same layering pattern as Agent Builder (global Role + per-topic detail).

sf-ai-agentforce-persona/CREDITS.mdLOW
68.3%

Tool passed security scan

sf-ai-agentforce-persona/references/persona-framework.mdLOW
62.7%

Tool passed security scan

Passed Files (4)Click to expand
sf-ai-agentforce-persona/README.mdOK
30.7%

Tool passed security scan

sf-ai-agentforce-persona/LICENSEOK
18.3%

Tool passed security scan

sf-ai-agentforce-persona/assets/persona-encoding-template.mdOK
13.4%

Tool passed security scan

sf-ai-agentforce-persona/assets/persona-template.mdOK
5.6%

Tool passed security scan

Audit Metadata
Max File Score
95%
Classification
KNOWN_SERVER_KNOWN_TOOL_LIST
Files Scanned
8
Files Flagged
4
Chunks Analyzed
17
Analyzed
Mar 13, 2026, 05:58 AM
Security Audit — runlayer — sf-ai-agentforce-persona