Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/doodledood/manifest-dev/sync-claude-code-pluginsnpx skills add doodledood/manifest-dev --skill sync-claude-code-pluginsgit clone --depth 1 https://github.com/doodledood/manifest-devWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00089 | $0.01616 |
| Opus 5 | $0.00044 | $0.00808 |
| Sonnet 5 | $0.00018 | $0.00323 |
| Haiku 4.5 | $0.00009 | $0.00162 |
Grade C, and why
sync-claude-code-plugins scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- **Symlinks look like directories to `cp`/`rm`/`find`**: a symlinked target overwritten by `cp -R` corrupts the linked plugin's source files; a symlinked directory deleted by `rm -rf` removes the link, not the plugin, b How it starts
The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.
User request: $ARGUMENTS
Sync prompt-engineering plugin components from a local sibling clone of claude-code-plugins into this repo's .claude/ directory. The plugin OWNS only the files it ships — other content in .claude/agents/ and .claude/skills/ (manifest-dev sync, KB skills, anything else) must be left alone.
Source & Target
| Role | Path |
|---|---|
| Source repo | ../claude-code-plugins (relative to this repo's root) |
| Source components | <source_repo>/claude-plugins/prompt-engineering/ |
| Target | .claude/ in this repo |
| Tracking file | .claude/.claude-code-plugins-sync.json |
Sync scope
| Component | Source dir | Target dir |
|---|---|---|
| Agents | agents/ |
.claude/agents/ |
| Skills | skills/ |
.claude/skills/ |
prompt-engineering ships no hooks today.
Territory model
Deletion invariant: only items in tracked (the previously-synced set) are eligible for removal when they disappear upstream. Items never in tracked are invisible — that's how project-local content stays safe.
The tracked set lives in .claude/.claude-code-plugins-sync.json:
{
"version": 1,
"last_synced_at": "ISO-8601 timestamp",
"agents": ["prompt-reviewer.md", "..."],
"skills": ["prompt-engineering", "..."]
}
First run (file missing): tracked is empty, no deletions happen, file is written at end.
Sync algorithm
Pre-flight: abort if <source_repo>/claude-plugins/prompt-engineering/ is missing — silent absence is a misconfigured path, not upstream removal. Do not delete on this signal. If the source is a clean git repo, git pull --ff-only first; surface a warning and proceed if pulling fails.
For each component (agents/skills):
- Copy every source item over its target path. Skip if target is a symlink. Skip retired items (see denylist below) even when an older upstream checkout still ships them.
- Re-apply local metadata after copying: every non-symlink synced skill's
SKILL.mdmust carrymetadata:/internal: truein its frontmatter — add it back when the upstream copy lacks it. Synced skills are repo-maintenance tooling in this repo; the flag keeps skill pickers and directory scrapers from listing them as product skills. - Delete items in
tracked − sourcefrom target. Skip if target is a symlink, doesn't exist, or is thesync-claude-code-pluginsskill itself. - Refresh
.claude/.claude-code-plugins-sync.jsonwith the current source listing minus the denylist.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 94 lines · 89 tokens per session scan C 16594fde4168
sync-claude-code-plugins is a skill published in the GitHub repository doodledood/manifest-dev (72 stars, last pushed 3d ago), licensed MIT. It adds 89 tokens to every session and 1,616 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
autopilot
Full autonomous execution from idea to working code.
deepinit
Deep codebase initialization with hierarchical AGENTS.md documentation.
graph
Deterministic orchestration graph runtime - declarative DAG pipelines with journal-based crash recovery.
ask
Process-first advisor routing for Claude, Codex, Gemini, Antigravity, Grok, or Cursor via omc ask, with artifact capture and no raw CLI assembly.
mcp-builder
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).
deep-interview
Socratic deep interview with mathematical ambiguity gating before explicit execution approval.