Empirical research into how agent platforms implement Agent Skill loading, management, and presentation. Benchmark skills and per-platform findings.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
First skill in a three-skill invocation chain for testing cross-skill invocation, depth limits, and language sensitivity. Use when asked to start an invocation chain.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Second skill in a three-skill invocation chain. Activated by invoke-alpha, activates invoke-gamma. Use when directed by invoke-alpha.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Third and final skill in a three-skill invocation chain. Activated by invoke-beta. Use when directed by invoke-beta.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill installed at the cross-client .agents/skills convention path rather than the platform's native skills directory. Use when asked to probe cross-client interop.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill whose name violates the consecutive-hyphen rule. Use when asked to probe double-hyphen name handling.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill whose name violates the lowercase-only rule. Use when asked to probe uppercase name handling.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Control twin of probe-allowed-tools with no allowed-tools field. Use when asked to probe allowed-tools control behavior.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing the experimental allowed-tools field. Use when asked to probe allowed-tools behavior.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
First of a pair of skills that reference each other, for testing circular invocation handling. Use when asked to probe circular dependencies.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Second of a pair of skills that reference each other, for testing circular invocation handling. Use when directed by probe-circular-alpha.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing name-collision precedence. This is the USER-scope variant. Use when asked to probe name collisions.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing how platforms handle the compatibility field. Use when asked to probe compatibility behavior.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing cross-scope dependency resolution. Install at project level; it references a skill expected at user level. Use when asked to probe cross-scope dependencies.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing deeply nested resource access and nested skill discovery. Use when asked to probe nesting behavior.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
A skill nested inside another skill's references directory. Used to test whether platforms discover skills inside other skills' directory trees.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill nested one level below the skills root inside a plain grouping directory (probe-group). Use when asked to probe grouped skill discovery.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing eager link resolution and path resolution behavior. Use when asked to probe link resolution.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing loading timing, resource enumeration, content presentation, and lifecycle behavior. Use when asked to probe loading behavior.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill whose compatibility field runs past the spec's 500-character limit. Use when asked to probe oversize compatibility handling.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill whose description deliberately runs past the spec's 1024-character limit, to test oversize field handling. Use when asked to probe oversize description handling. The head marker SANDERLING-GNEISS-1010 sits near the start of this description and a matching tail marker sits at the very end, so testers…
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Use when: asked to probe malformed yaml parsing behavior.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing how platforms handle edge-case YAML values in the metadata frontmatter field. Use when asked to probe metadata value handling.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill whose frontmatter name deliberately differs from its directory name (probe-mismatch-dir). Use when asked to probe name mismatch handling.
agent-ecosystem/agent-skill-implementation
Skill Claude CodeCodex
Benchmark skill for testing how platforms handle references to skills that don't exist. Use when asked to probe missing dependency behavior.