Agent Workspace

Pack comparison

Two packs, side-by-side. Merged comparisons, shared shape, and diff highlights in one view.

ACommunitysecurityv0.1.0Recommended
Seven Safety Layers

seven-safety-layers

Defense-in-depth for tool execution. Deny > ask > allow. All 7 layers, in order, with honest failure modes.

npx attrition-sh pack install seven-safety-layers

Token budget

Pass rate

Avg tokens

Publisher

Agent Workspace

claude-codecursorcodex
BCommunitysecurityv0.1.0Recommended
Injection Surface Audit

injection-surface-audit

Every agent product ships injection surfaces. Audit them before an attacker does.

npx attrition-sh pack install injection-surface-audit

Token budget

Pass rate

Avg tokens

Publisher

Agent Workspace

claude-codecursorcodex-cliany-agent-harness

What both packs have in common

Overlap across canonical pattern, compatibility, tags, and required packs.

claude-codecursor
security

Head-to-head claims from both packs

Each row is attributed to the pack that authored it. The winner column is normalised to this compare view (A / B / Tie).

SourceAlternativeAxisWinnerNote
Ainjection-surface-auditmaintainabilityTieThis pack is the architecture; injection-surface-audit is the checklist. Use the architecture for design decisions, the checklist for per-release audits. Complementary, not substitutes.
Aturn-execution-pipelinecomplexityAlternativeTurn pipeline describes the outer 9-step loop; this pack expands step 7 (permission gate) into its 7-layer decomposition. Pipeline is simpler because it treats the gate as one step; safety layers are the hard part.
Aowasp-llm-top10accuracyAOWASP LLM Top 10 is a vocabulary for LLM-app vulnerabilities; this pack is a specific architecture. Use OWASP to name the risks, this pack to verify the layers.
Bowasp-llm-top10maintainabilityBOWASP LLM Top-10 is a vocabulary; this pack is an actionable checklist tied to specific code patterns. Use together: OWASP for framing, this pack for line-level audit.
Bllm-guardrails-middlewarecomplexityAlternativeRuntime guardrail middleware (NeMo, Guardrails AI) adds automated filtering — lower manual effort, adds a dependency. This pack is zero-runtime and targets design-time holes. Layered defence uses both.
Bred-team-engagementaccuracyAlternativeA professional red-team engagement finds novel classes a checklist can't. Use this pack monthly; commission a red-team annually.

What each pack brings that the other doesn't

Unique coverage and any measurable gap between the two.

Comparisons not in B

injection-surface-auditturn-execution-pipeline

Compatibility A-only

codex

Tags A-only

permissionsdefense-in-depthclaude-code-internalsdive-into-claude-codessrf-and-sandbox

Comparisons not in A

llm-guardrails-middlewarered-team-engagement

Compatibility B-only

codex-cliany-agent-harness

Tags B-only

prompt-injectionssrfauditowasp-llmsupply-chain