Free tools Windows power users keep installed
One-click scans. No signup required.
If Claude Code returns Unknown skill: probe-api for a skill stored inside a monorepo package, it may not have discovered that package’s skill yet. In a small test on Claude Code 2.1.285, using the dedicated Read tool on a file in the package triggered discovery; reading the same file with Bash cat did not. That is a version-specific result, not a guarantee about every tool or release.
Why Claude Code can say “Unknown skill”
A skill inside a package’s .claude/skills/ directory may not be available at session start. Anthropic’s Claude Code skills documentation says nested skills load when Claude first reads or edits a file in that subdirectory, then remain available for the rest of the session. A Skill tool call made before discovery can therefore return Unknown skill.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
The key uncertainty is what counts as Claude “reading” a file in practice. The available reports suggest discovery can depend on the tool path, so a shell command that visibly prints a file’s contents may not have the same effect as Claude Code’s dedicated file tool.
What the Read-versus-cat test found
Rulestack’s report describes an experiment run on 2026-09-30 with Claude Code 2.1.285. Its principal test used 20 headless runs with Opus 5.5; a further ten runs repeated five configurations using Sonnet 5.5. These are small, configuration-specific results, not a broad benchmark.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
| Test setup | Reported result |
|---|---|
Bash cat on a package file |
0 of 2 runs loaded the nested skill; a subsequent Skill call returned Unknown skill. |
Dedicated Read tool on the package file |
6 of 6 runs loaded the nested skill. |
| Pre-discovery Skill calls | All ten reported Opus attempts returned Unknown skill; the report also describes pre-discovery failures in the Sonnet rerun. |
Those counts come from Rulestack’s test report. Its partial Sonnet 5.5 rerun matched the Read-versus-cat outcome. The sample does not establish that every Read call will trigger discovery, or that cat can never do so in another version or setup.
How to make a package skill available
Use Read on a file inside the package
If the skill is nested under a package subtree, ask Claude to use the dedicated Read tool on a file in that subtree before calling the skill. In the 2.1.285 test, that route loaded the skill consistently across the six reported Read runs. Treat it as a practical workaround supported by that test, rather than a universal product guarantee.
Rank #2
Add the directory earlier when the skill is needed from the start
Anthropic documents /add-dir as a way to add a directory so its skills are available earlier; the documentation specifies Claude Code v2.1.257 or later. See the skills documentation for the current command guidance.
Rulestack also found that starting a session inside the package loaded its skill at startup. Its tested alternative was the --add-dir launch flag, not the /add-dir command. The report cautions that starting inside a package can change which shared settings are loaded, so that approach may affect more than skill discovery.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsDo not assume every file-reading route triggers discovery
A separate open report, Anthropic Claude Code GitHub issue #93403, describes nested skills and subdirectory instruction files appearing after Read, but not after Bash cat, sed, or grep, nor after the dedicated Grep tool. That reporter used Claude Code 2.1.267 on Windows. It is an individual issue report, not an official finding that the behavior is universal or resolved.
The issue also points to a workflow tension: an auto mode that directs Claude toward Bash-first work may steer it away from the file tools that the reporter found triggered discovery. Anthropic’s tools reference documents the available tools, but neither that reference nor these reports prove how every access path interacts with nested-skill loading in every release.
What to do if the skill name collides
If a nested skill has the same name as a root-level skill, check the directory-qualified skill name Claude Code surfaces after discovery. Anthropic’s skills documentation describes qualified variants for nested skills that share a name with a root skill. In Rulestack’s collision test, Claude surfaced and used the directory-qualified nested deploy variant in both runs. That is evidence of what happened in those runs, not a promise that Claude will always select the intended variant automatically.
How much confidence to put in the result
The practical distinction is clear in the cited reproduction: on Claude Code 2.1.285, a dedicated Read on a package file triggered skill discovery in 6 of 6 runs, while Bash cat did not in 0 of 2. A separate report on version 2.1.267 describes similar non-triggering behavior for other tools. The evidence is limited to those setups and counts; it does not establish prevalence among users or behavior in later releases.
Rulestack also estimated that a skill listed at session start added 60 input tokens to the first request in its tested configuration, while a nested skill’s later arrival added about 140 tokens by the author’s calculations. The report says the Sonnet rerun showed the same listing costs. These are configuration-specific token observations, not general Claude Code overhead figures.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
