CtrlK
BlogDocsLog inGet started
Tessl Logo

agent-architecture-audit

Full-stack diagnostic for agent and LLM applications. Audits the 12-layer agent stack for wrapper regression, memory pollution, tool discipline failures, hidden repair loops, and rendering corruption. Produces severity-ranked findings with code-first fixes. Essential for developers building agent applications, autonomous loops, or any LLM-powered feature. Use when an agent or LLM feature misbehaves and the failing layer is unknown, or before shipping an agent stack.

67

Quality

81%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

67%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, actionable audit workflow with executable search commands, a clear phased sequence, and a concrete report schema. Its main weakness is redundancy across three overlapping failure taxonomies and content length that could benefit from splitting detail into reference files.

Suggestions

Consolidate the three overlapping failure framings (the 12-layer table, Common Failure Patterns, and Quick Diagnostic Questions) into a single canonical taxonomy, cross-referencing rather than restating each.

Move the full JSON Report Schema and the detailed per-pattern Symptom lists into a reference file (e.g. references/report-schema.md), keeping SKILL.md as an overview that links out.

Add an explicit validation checkpoint in the Audit Workflow (e.g., a Phase 3.5 'Re-verify each finding's evidence_refs and confidence before ranking') to turn the implicit quality gates into a feedback loop.

DimensionReasoningScore

Conciseness

Mostly efficient tables and lists with no basic-concept padding, but the same failure space is covered three times over — the 12-layer table, the 5 Common Failure Patterns, and the 7 Quick Diagnostic Questions — creating redundant taxonomies that could be consolidated.

3 / 5

Actionability

Provides copy-paste-ready `rg` search commands in Phase 2, a concrete ordered fix list in Phase 4, and a full JSON report schema; minor gaps remain where fix strategies are stated as principles (e.g., "Tighten memory admission — user corrections > agent assertions") rather than executable steps.

4 / 5

Workflow Clarity

A clear four-phase sequence (Scope → Evidence Collection → Failure Mapping → Fix Strategy) with evidence_refs and a 0.0–1.0 confidence field acting as soft checkpoints; not a 5 because there is no explicit validate-and-revisit feedback loop for findings.

4 / 5

Progressive Disclosure

Well-organized into clearly headed sections and self-contained with no broken or nested references; not a 5 because at ~250 lines some content (the detailed failure-pattern writeups, the full report schema) could be externalized into reference files rather than inlined.

4 / 5

Total

15

/

20

Passed

Description

96%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific description that clearly states what the skill does and when to use it, with concrete failure-mode triggers and comprehensive keyword coverage. The only soft spot is minor overlap risk with sibling agent skills, which the body (not the description) resolves via a 'Do not use for' section.

DimensionReasoningScore

Specificity

Lists multiple concrete actions and a comprehensive set of failure modes — "Audits the 12-layer agent stack for wrapper regression, memory pollution, tool discipline failures, hidden repair loops, and rendering corruption" plus "Produces severity-ranked findings with code-first fixes".

5 / 5

Completeness

Explicitly answers both what (audits the 12-layer stack, produces severity-ranked findings) and when ("Use when an agent or LLM feature misbehaves and the failing layer is unknown, or before shipping an agent stack") with concrete trigger phrases.

5 / 5

Trigger Term Quality

Natural trigger phrases users would say are present with synonyms — "agent or LLM feature misbehaves", "failing layer is unknown", "before shipping an agent stack"; comprehensive domain keyword coverage.

5 / 5

Distinctiveness Conflict Risk

Carves a clear niche (12-layer stack audit, wrapper regression, hidden repair loops) but "agent applications" framing has minor overlap risk with related agent skills like agent-eval or agent-introspection-debugging; not a 5 because the description alone does not fully disambiguate from those neighbors.

4 / 5

Total

19

/

20

Passed

Validation

87%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation14 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

14

/

16

Passed

Repository
affaan-m/ECC
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.