* feat(mcp): pin tool definitions in mcp check --live and warn when they change An Agent reads a server's tool names and descriptions as instructions, and a server can change them after the user reviewed it (a "rug pull"). The MCP spec 2026-07-28 local-server security guide notes the protocol does nothing about this and advises clients that re-prompt on a changed definition; most clients do not. `mcp check --live` now records one SHA-256 per tool over name, description and inputSchema the first time a server answers, in mcp/tool-pins.json of the state directory, keyed by scope and server. Later probes report a changed or new tool as a `tool-change` warning and a removed one as info, until `--accept-tools` (only with --live) records the current definitions. Without --live the pins are never read or written. An unreadable pin file is never rewritten, since it may hold other servers' pins. The dashboard words both warnings in every locale. * fix(mcp): keep schema numbers exact, name the scope to accept in, and log acceptance Codex review on #560: - Tool definitions were decoded through float64 before hashing, so a schema bound changed from 9007199254740992 to 9007199254740993 hashed the same and raised no warning. Numbers now keep their literal text. - The suggested accept command left out -g/-p; run from inside a project after a global check, it would accept the project's server instead. It now names the scope of the check that warned. - Accepting changed tool definitions approves a possible rug pull, but left no record. --accept-tools now writes an operations log entry with the servers it accepted. * fix(mcp): log a tool acceptance only when its baseline was saved Codex review on #560: when comparing succeeded but writing tool-pins.json failed (lock held, state directory unwritable), the report kept its "accepted" finding, so the operations log listed the server as accepted with status ok although its old pins remained. Such a server is now logged under `failed` and the entry's status is error. * fix(mcp): log first pins as accepted and report errors in the accept entry Codex review on #560: - A check that accepted one server but found an error on another still logged status ok, while the command itself exited 1. A server with an error is now listed under `failed` and the entry's status is error. - --accept-tools on a server's first probe records its baseline, but the finding says "pinned", so the entry listed nothing as accepted. A first pin counts as accepted now. * chore(release): request version 0.25.11 The MCP pin checks (#559, #560) and the backup staging fix (#558) ship as 0.25.11 instead of the computed 0.26.0, per maintainer decision. Release-As: 0.25.11
16 KiB
sidebar_position
| sidebar_position |
|---|
| 3 |
File Structure
Directory layout and file locations for skillshare.
Overview
~/.config/skillshare/ # XDG_CONFIG_HOME
├── config.yaml # Configuration file
├── audit-rules.yaml # Custom audit rules (optional)
├── mcp.yaml # MCP servers, if sources.mcp points here (optional)
├── skills/ # Skills source (skills + metadata)
│ ├── .metadata.json # Installed skill metadata (auto-managed)
│ ├── .skillignore # Optional: exclude skills from sync
│ ├── my-skill/ # Regular skill
│ │ ├── SKILL.md # Skill definition (required)
│ ├── code-review/ # Another skill
│ │ └── SKILL.md
│ └── _team-skills/ # Tracked repository
│ ├── .git/ # Git history preserved
│ ├── frontend/
│ │ └── ui/
│ │ └── SKILL.md
│ └── backend/
│ └── api/
│ └── SKILL.md
├── agents/ # Agents source (single .md files)
│ ├── .agentignore # Optional: exclude agents from sync
│ ├── reviewer.md # Agent file
│ └── auditor.md # Another agent
├── rules/ # Extras source (if configured)
│ ├── coding.md
│ └── testing.md
└── commands/ # Extras source (if configured)
└── deploy.md
~/.local/share/skillshare/ # XDG_DATA_HOME
├── backups/ # Backup directory
│ ├── 2026-01-20_15-30-00/
│ │ ├── claude/ # Skills backup for claude
│ │ ├── claude-agents/ # Agents backup for claude
│ │ └── cursor/
│ └── 2026-01-19_10-00-00/
│ └── claude/
└── trash/ # Uninstalled skills/agents (7-day retention)
├── my-skill_2026-01-20_15-30-00/
│ └── SKILL.md
└── old-skill_2026-01-19_10-00-00/
└── SKILL.md
~/.local/state/skillshare/ # XDG_STATE_HOME
├── logs/ # Operation logs (JSONL)
│ ├── operations.log # install, sync, update, etc.
│ └── audit.log # Security audit scans
├── mcp/ # MCP sync state (auto-managed)
│ ├── state.json # Which native entries Skillshare owns
│ ├── tool-pins.json # Tool definition hashes recorded by mcp check --live
│ └── backups/ # Agent files before each write (newest 20 per file)
└── plugins/ # Reviewed local copies of plugin sources
~/.cache/skillshare/ # XDG_CACHE_HOME
├── version-check.json # Version check cache (24h TTL)
└── ui/ # Web UI dist cache
└── 0.13.0/ # Per-version cached assets
├── index.html
└── assets/
Configuration File
Location
~/.config/skillshare/config.yaml
Override with XDG:
XDG_CONFIG_HOME=/custom/path → /custom/path/skillshare/config.yaml
Windows default:
%AppData%\skillshare\config.yaml
Contents
# yaml-language-server: $schema=https://raw.githubusercontent.com/runkids/skillshare/main/schemas/config.schema.json
source: ~/.config/skillshare/skills
agents_source: ~/.config/skillshare/agents # Optional; defaults to <source parent>/agents
mode: merge
targets:
claude:
path: ~/.claude/skills
agents: # Optional; enables agent sync for this target
path: ~/.claude/agents
cursor:
path: ~/.cursor/skills
ignore:
- "**/.DS_Store"
- "**/.git/**"
See Configuration for full reference.
Metadata File
Location
~/.config/skillshare/skills/.metadata.json
Stores metadata about installed and tracked skills. Lives inside the source directory so it can be synced via git for multi-machine setups. Auto-managed by install, uninstall, and update — don't edit manually.
Contents
{
"skills": [
{
"name": "pdf",
"source": "anthropics/skills/skills/pdf"
},
{
"name": "_team-skills",
"source": "github.com/team/skills",
"tracked": true
}
]
}
Each entry records the skill name and its install source. Tracked repos (prefixed with _) include the full repository URL for update and check operations.
Source Directory
Location
~/.config/skillshare/skills/
Windows:
%AppData%\skillshare\skills\
Structure
skills/
├── .metadata.json # Centralized skill metadata (auto-managed)
├── skill-name/ # Skill directory
│ ├── SKILL.md # Required: skill definition
│ ├── examples/ # Optional: example files
│ └── templates/ # Optional: code templates
├── frontend/ # Category folder (via --into or manual)
│ └── react-skill/ # Skill in subdirectory
│ └── SKILL.md # Synced as frontend__react-skill
└── _tracked-repo/ # Tracked repository
├── .git/ # Git history
└── ... # Skill subdirectories
Skill Files
SKILL.md (Required)
The skill definition file:
---
name: skill-name
description: Brief description
---
# Skill Name
Instructions for the AI...
See Skill Format for details.
.skillignore (Optional)
Excludes skills from discovery. Supports two locations:
Repo-level — at the root of a tracked skill repository. Affects install discovery and all post-install commands (doctor, status, list, sync, etc.):
# Hide vendored packages from discovery
.venv
node_modules
# Exclude internal tooling
validation-scripts
prompt-eval-*
Source-root — at your source directory root (~/.config/skillshare/skills/.skillignore). Applies globally to all skills (tracked and non-tracked):
# Temporarily mute a skill
my-experimental-skill
# Exclude all drafts
draft-*
Uses gitignore syntax — one pattern per line. Supports * (single segment), ** (any depth), ?, [abc] (character class), !pattern (negation), /pattern (anchored), pattern/ (directory-only), and \#/\! (escaped literals). Lines starting with # are comments. A group name like internal-tools excludes all skills under that directory; internal-tools/helper excludes only a specific skill. Both layers apply — if either matches, the skill is excluded.
:::tip
.skillignore is one of three filtering layers. See Filtering Skills for all scenarios including per-target filters and SKILL.md targets.
:::
.skillignore.local (Optional)
A local-only override file that works alongside .skillignore. Place it in the same directory as a .skillignore (source root or tracked repo root). Patterns from .skillignore.local are appended after .skillignore, so negation patterns (!pattern) can override the base file:
# The repo's .skillignore blocks private-*, but I need my own
!private-mine
This file should not be committed to version control — add it to .gitignore. It exists so that a repo consumer can locally override the repo maintainer's .skillignore without modifying it.
When active, sync -v, status, and doctor display a .local active indicator.
Agent Files
Agents are a separate resource kind from skills. They live in a sibling source directory and are synced to agent-capable targets (Claude, Cursor, Augment, OpenCode).
Agent source directory
~/.config/skillshare/agents/ # Global mode
.skillshare/agents/ # Project mode
The agents source is created automatically by skillshare init alongside skills/. You can override the global location with the agents_source config field; project mode always uses .skillshare/agents/.
Agent file format
Each agent is a single Markdown file with frontmatter:
---
name: reviewer
description: Reviews pull requests for security and style issues.
---
# Reviewer
Instructions for the AI agent...
Agent filenames must use only a-z, 0-9, _, -, .. Unlike skills, agents are single files — they do not contain subdirectories.
See Agents for the full file format and discovery rules.
.agentignore (Optional)
Excludes agents from sync. Lives at the agents source root:
# Hide drafts
draft-*
# Disable a specific agent
experimental-reviewer
Uses gitignore syntax. The same pattern rules as .skillignore apply (*, **, !negation, # comments, etc.). Disabled agents remain in the source directory but are excluded from sync.
skillshare disable <agent> and skillshare enable <agent> add/remove entries automatically.
.agentignore.local (Optional)
A local-only override file (same pattern as .skillignore.local). Place it next to .agentignore. Patterns are appended after .agentignore, so !negation patterns can re-enable agents the base file disables. Should not be committed to version control.
Backup Directory
Location
~/.local/share/skillshare/backups/
Structure
backups/
└── <timestamp>/ # YYYY-MM-DD_HH-MM-SS
├── claude/ # Backup of target
│ ├── skill-a/
│ └── skill-b/
└── cursor/
└── ...
Backups are created:
- Automatically before
syncandtarget remove - Manually via
skillshare backup
Trash Directory
Location
~/.local/share/skillshare/trash/
Project mode:
<project>/.skillshare/trash/
Structure
trash/
└── <skill-name>_<timestamp>/ # skill-name_YYYY-MM-DD_HH-MM-SS
├── SKILL.md
└── ... # All original files preserved
Trashed skills are:
- Created by
skillshare uninstall - Retained for 7 days, then automatically cleaned up
- Named with the original skill name and a timestamp
Log Directory
Location
~/.local/state/skillshare/logs/
Project mode:
<project>/.skillshare/logs/
Target Directories
Targets are AI CLI skill directories. After sync, they contain symlinks (or copies) to source.
Merge mode
Each skill is symlinked individually. A manifest tracks managed skills for orphan cleanup:
~/.claude/skills/
├── my-skill -> ~/.config/skillshare/skills/my-skill
├── code-review -> ~/.config/skillshare/skills/code-review
├── local-only/ # Not symlinked (user-created, preserved)
└── .skillshare-manifest.json # Tracks managed skills
Copy mode
Each skill is copied as real files. The manifest tracks checksums for incremental sync:
~/.cursor/skills/
├── my-skill/ # Real files (copied from source)
├── code-review/ # Real files
├── local-only/ # User-created, preserved
└── .skillshare-manifest.json # Tracks managed skills + checksums
Symlink mode
Entire directory is symlinked:
~/.claude/skills -> ~/.config/skillshare/skills/
Copies in place of file links (Windows)
On Windows without Developer Mode, agent targets and directory extras in merge mode get copies instead of file links. The target directory then has a .skillshare-manifest.json that records a checksum for each copy, so later syncs can update or prune them and leave your own files alone:
~/.claude/agents/
├── reviewer.md # Copy of the source agent
├── local-agent.md # User-created, preserved
└── .skillshare-manifest.json # Tracks copied agents + checksums
A single-file extra (such as a shared AGENTS.md) records its copy in skillshare's extras backup folder instead of the target directory.
Tracked Repositories
Tracked repos (installed with --track) preserve git history:
_team-skills/
├── .git/ # Git preserved
├── frontend/
│ └── ui/
│ └── SKILL.md
└── backend/
└── api/
└── SKILL.md
Naming conventions
_prefix: tracked repository__in flattened name: path separator
In source:
_team-skills/frontend/ui/SKILL.md
In target (flattened):
_team-skills__frontend__ui/SKILL.md
Platform Differences
:::tip XDG Base Directory
skillshare respects the XDG Base Directory Specification. Override base directories with XDG_CONFIG_HOME, XDG_DATA_HOME, XDG_STATE_HOME, and XDG_CACHE_HOME.
See Environment Variables for details. :::
macOS / Linux
| Item | Path |
|---|---|
| Config | ~/.config/skillshare/config.yaml |
| Metadata | ~/.config/skillshare/skills/.metadata.json |
| Skills source | ~/.config/skillshare/skills/ |
| Agents source | ~/.config/skillshare/agents/ |
| Backups | ~/.local/share/skillshare/backups/ |
| Trash | ~/.local/share/skillshare/trash/ |
| Logs | ~/.local/state/skillshare/logs/ |
| Version cache | ~/.cache/skillshare/version-check.json |
| UI cache | ~/.cache/skillshare/ui/{version}/ |
| Link type | Symlinks |
Windows
| Item | Path |
|---|---|
| Config | %AppData%\skillshare\config.yaml |
| Metadata | %AppData%\skillshare\skills\.metadata.json |
| Skills source | %AppData%\skillshare\skills\ |
| Agents source | %AppData%\skillshare\agents\ |
| Backups | %AppData%\skillshare\backups\ |
| Trash | %AppData%\skillshare\trash\ |
| Logs | %AppData%\skillshare\logs\ |
| Version cache | %AppData%\skillshare\version-check.json |
| UI cache | %AppData%\skillshare\ui\{version}\ |
| Link type | NTFS Junctions for folders; symlinks for single files (need Developer Mode, otherwise copied) |
XDG Base Directory Layout
skillshare follows the XDG Base Directory Specification on Unix systems:
| XDG Variable | Default Path | skillshare Uses For |
|---|---|---|
XDG_CONFIG_HOME |
~/.config |
skillshare/config.yaml, skillshare/skills/ (includes .metadata.json), skillshare/agents/ |
XDG_DATA_HOME |
~/.local/share |
skillshare/backups/, skillshare/trash/ |
XDG_STATE_HOME |
~/.local/state |
skillshare/logs/ |
XDG_CACHE_HOME |
~/.cache |
skillshare/ui/ (downloaded web dashboard) |
Windows Paths
| Purpose | Path |
|---|---|
| Config + Skills | %AppData%\skillshare\ |
| Data (backups, trash) | %AppData%\skillshare\ |
| State (logs) | %AppData%\skillshare\ |
| Cache (UI) | %AppData%\skillshare\ |
Migration Note
If upgrading from a version before the XDG split, skillshare automatically migrates data from the old location (~/.config/skillshare/) to the correct XDG directories on first run.
Related
- Configuration — Config file details
- Skill Format — SKILL.md format
- Agents — Agent file format and discovery
- Tracked Repositories — Tracked repos