Skill source(s). If omitted, installs from skillx.toml
Flags
Flag
Type
Default
Description
--agent
string
auto-detect
Target a specific agent (conflicts with --all)
--all
bool
false
Install to all detected agents (conflicts with --agent)
--scope
string
global
Injection scope: global or project
--no-cache
bool
false
Force re-fetch, skip cache
--skip-scan
bool
false
Skip security scan
--yes
bool
false
Auto-confirm WARN level risks
--no-save
bool
false
Don’t save to skillx.toml
--dev
bool
false
Install as dev dependency
--prod
bool
false
Only install production dependencies, skip dev
--prune
bool
false
Remove installed skills not in skillx.toml
Examples
Terminal window
skillxinstallgithub:org/skills/pdf
skillxinstall--all--devgithub:org/skills/testing
skillxinstall# from skillx.toml
skillxinstall--prod--prune# production only, prune extras
skillx uninstall
Terminal window
skillxuninstall<name...> [options]
Arguments
Argument
Position
Required
Description
names
1+
Yes
Skill name(s) to uninstall
Flags
Flag
Type
Default
Description
--agent
string
—
Only remove from a specific agent
--keep-in-toml
bool
false
Keep the entry in skillx.toml
--purge
bool
false
Also remove cached files
Examples
Terminal window
skillxuninstallpdf-processing
skillxuninstallformatter--agentcursor
skillxuninstallold-skill--purge
skillx list
Terminal window
skillxlist [options]
Flags
Flag
Type
Default
Description
--agent
string
—
Filter by agent
--scope
string
all
Filter by scope: project, global, or all
--json
bool
false
Output as JSON (to stdout)
--outdated
bool
false
Check for available updates
Examples
Terminal window
skillxlist
skillxlist--agentclaude-code
skillxlist--json|jq'.[].name'
skillxlist--outdated
skillx update
Terminal window
skillxupdate [names...] [options]
Arguments
Argument
Position
Required
Description
names
1+
No
Skill name(s) to update. If omitted, checks all
Flags
Flag
Type
Default
Description
--agent
string
—
Only update for a specific agent
--dry-run
bool
false
Show what would be updated without applying
--skip-scan
bool
false
Skip security scan
--yes
bool
false
Auto-confirm (skip update prompt)
Examples
Terminal window
skillxupdate
skillxupdate--dry-run
skillxupdatepdf-processing--yes
skillx init
Terminal window
skillxinit [options]
Flags
Flag
Type
Default
Description
--from-installed
bool
false
Pre-populate with currently installed skills
Examples
Terminal window
skillxinit
skillxinit--from-installed
skillx scan
Terminal window
skillxscan<source> [options]
Arguments
Argument
Position
Required
Description
source
1
Yes
Skill source to scan
Flags
Flag
Type
Default
Description
--format
string
text
Output format: text, json, or sarif
--fail-on
string
danger
Exit code threshold: info, warn, danger, block
Output Routing
text format: output to stderr
json / sarif format: output to stdout (pipeable)
Examples
Terminal window
skillxscan./my-skill
skillxscan--formatjsongithub:org/repo/skill
skillxscan--formatsarif./my-skill
skillxscan--fail-onwarn./my-skill
skillx agents
Terminal window
skillxagents [options]
Flags
Flag
Type
Default
Description
--all
bool
false
Show all known agents, including undetected
Examples
Terminal window
skillxagents
skillxagents--all
skillx info
Terminal window
skillxinfo<source>
Arguments
Argument
Position
Required
Description
source
1
Yes
Skill source to inspect
Show metadata parsed from the skill’s SKILL.md frontmatter.
Examples
Terminal window
skillxinfo./my-skill
skillxinfogithub:org/repo/skill
skillx cache
Terminal window
skillxcache<subcommand>
Subcommands
Subcommand
Description
ls
List all cached skills with source and timestamp
clean
Remove all cached skill entries
Examples
Terminal window
skillxcachels
skillxcacheclean
Environment Variables
Variable
Description
GITHUB_TOKEN
GitHub personal access token for private repositories
Exit Codes
Code
Applies To
Meaning
0
All commands
Success
1
All commands
Error or threshold exceeded
For skillx scan, exit code 1 means findings were found at or above the --fail-on threshold. For skillx run, exit code 1 means the run could not complete (source resolution failed, scan blocked, agent error, etc.).