cache-audit

Audit your Claude Code setup against prompt caching best practices. Checks ordering, tool stability, dynamic content handling, and hook injection patterns. Returns a scored report with specific fixes.

Quality

68%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/agentic-harness/cache-audit/SKILL.md

Quality

Content

71%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 highly actionable, well-sequenced audit procedure with concrete files, patterns, thresholds, and a specified output format. Its weaknesses are redundancy — the six rules are repeated across three sections — and a progressive-disclosure gap where the existing references/checklist.md bundle file is never linked from the body.

Suggestions

Link the existing bundle file explicitly, e.g. a "**Quick checklist**: See [checklist.md](references/checklist.md)" line, so the reference is discoverable rather than orphaned.

Consolidate the rules into one place: keep the detailed checks and replace the cheatsheet table and Anti-Patterns section with a pointer to references/checklist.md, cutting roughly a third of the body.

Add a one-line method for measuring dynamic injection size in Check 5 (e.g. "wc -c the hook output or count the <system-reminder> block") so the size verification is as executable as the rest.

Dimension	Reasoning	Score
Conciseness	The checks are operational rather than tutorial-style, but the six rules are stated three times (detailed checks, the "Quick Cheatsheet" table, and the "Anti-Patterns" section), and Check 6 concedes "Claude Code handles it correctly by default" — content Claude already knows. Mostly efficient, but the repetition and Check 6 could be tightened, matching the anchor for 3 rather than 4.	3 / 5
Actionability	Each check names exact files to read (~/.claude/settings.json, CLAUDE.md hierarchy, hook files), exact patterns to match (hookSpecificOutput.additionalContext, defer_loading: true), numeric thresholds (<2k / 2k–10k / >10k chars), a copy-paste bash fix for oversized git status injection, and a fully specified report template. It is not a 4 because the common cases are covered end-to-end with executable detail.	5 / 5
Workflow Clarity	The sequence is unambiguous (read config files → run checks 1–6 → score each with explicit PASS/WARNING/FAIL criteria → emit the templated report), and per-check scoring criteria act as checkpoints. It falls short of 5 because some verification methods are left implicit — e.g. "Estimate the size of dynamic content injected per turn" gives thresholds but no command or method to measure the injection size.	4 / 5
Progressive Disclosure	The body has clear section headers, but the one bundle file (references/checklist.md) is never referenced or linked from SKILL.md — an orphaned reference that a reader cannot discover — and supplementary material (the cheatsheet table, anti-patterns) is inlined instead of split out. This matches the anchor for 3: structure exists, but references present are not clearly signaled.	3 / 5
	Total	15 / 20 Passed

Description

66%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 specific, action-oriented description with a clear niche and decent trigger terms. Its main weakness is the absence of any 'when to use' guidance, which both caps completeness and leaves natural trigger phrases (cache misses, breaking the cache) out of the description itself.

Suggestions

Add an explicit trigger clause to the description, e.g. "Use when cache hit rates are lower than expected, after changing the system prompt or tool list, or when the user asks about caching or token costs."

Include the natural trigger variations users actually say — "cache misses", "cache hit rate", "breaking the cache" — so the description surfaces on those queries without relying on the body's trigger line.

Mention the remaining check areas (model switching, dynamic content size thresholds) so the description's coverage matches the skill's actual scope.

Dimension	Reasoning	Score
Specificity	"Checks ordering, tool stability, dynamic content handling, and hook injection patterns" and "Returns a scored report with specific fixes" list several concrete actions with a defined output artifact. It stays at 4 rather than 5 because coverage has minor gaps (e.g., model switching and dynamic-content size are checks in the body but absent from the description).	4 / 5
Completeness	The "what" is clear (audits the setup against caching best practices, checks four areas, returns a scored report), but there is no "Use when..." clause or equivalent explicit trigger guidance in the description. Per the judging guidelines, a missing 'Use when' clause caps completeness at 3, and it is not a 2 because the 'what' is concrete rather than vague.	3 / 5
Trigger Term Quality	"Audit", "prompt caching", and "cache" are natural terms a user would say, giving good keyword coverage. A few natural variations are missing ("cache misses", "hit rate", "am I breaking the cache"), which appear only in the body's trigger line, not the description.	4 / 5
Distinctiveness Conflict Risk	"Prompt caching" scoping gives it a clear niche that few skills would claim, and the audit framing is distinctive. It is not a 5 because the description could overlap slightly with general Claude Code configuration or performance-tuning skills.	4 / 5
	Total	15 / 20 Passed

Validation

100%

Warnings & errors only

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

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: pantheon-org/tekhne
Path: skills/agentic-harness/cache-audit/SKILL.md
Commit: a1083f4

Reviewed: 2 hours ago

Table of Contents

Discovery Implementation Validation

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.