a
Safe
Adversarial Eval Scorecard Agent
Use this agent when an internal research team needs to design adversarial evaluation suites for search or research agents and produce reproducible JSON, Markdown, and CSV scorecards.
4 skills1 agent1 hook1 memory1 mcp1 command
STRUCTURE MAP
What's inside
Memory layer
adversarial-eval-scorecard-agent-memory
2.5KB
Agents
adversarial-eval-scorecard-agent
3 tools
Skills
i-am-an-openai-caliber
research <topic>
Hooks
PreToolUse
Bash
MCP servers
google-drive
stdio
Commands
/i-am-an-openai-caliber
2 relationships detected · agents → skills · hooks → mcp · commands → skills
WHAT THIS AI TEAM DOES
In one paragraph
Adversarial Eval Scorecard Agent contains 4 skills, 1 agent, 1 automated hook, and 1 MCP server. Atlas inferred likely themes from names and descriptions: research, design, and support.
BEST FOR
Audience
- developers browsing the repository structure
- maintainers documenting an agent or skill library
- reviewers deciding which files to inspect next
BEFORE YOU USE IT
Prerequisites
- CLAUDE.md was detected and should be read before using the stack
- AGENTS.md was detected and should be read before using the stack
- 1 MCP server detected in config
- memory files were detected and may affect agent behavior
EXPECT
Outputs
- an inventory of detected skills and their source paths
- an inventory of detected agents and declared tool hints
- 1 detected slash command
BE CAREFUL WITH
Permissions
- some detected files mention file write operations
Generated · source: deterministic
Access Summary
Facts inferred from file patterns and permission hints. This is not a quality rating.
read-only scanno repository uploadpattern-based
File accessscoped
Networkno
External APIsoptional
SecretsNone detected
Detected signals
- No exposed secret values were detected in scanned assets.
- Some detected assets can invoke shell commands or helper scripts.
- Some detected assets mention file write operations.
- Project policy defines 2 allow rule(s).
SECURITY REVIEW
Cleared — safe to publish
No exposed private keys, unsafe code, or over-broad permissions were found. The agent ships with a verified-safe badge.
0 critical0 warning0 info
Passes checks for exposed private keys, unsafe command access, outside data sharing, and over-broad permissions.
Your turn — review a ZIP or GitHub repo
Agentlas reads a ZIP or public GitHub repo, including its instructions, skills, automations, and memory — and gives you a clean profile like this one.
SHARE KIT
Make this profile travel
Download social PNG/SVG assets or copy a README badge that links back to this install page.
Badge = install previewQR = public profile linkREADME badge = source users can re-check