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/devhimanshulohani/claude-agent-manager/agentnpx skills add devhimanshulohani/claude-agent-manager --skill agentgit clone --depth 1 https://github.com/devhimanshulohani/claude-agent-managerWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/devhimanshulohani/claude-agent-manager/agent)<a href="https://agentmods.dev/skills/devhimanshulohani/claude-agent-manager/agent"><img src="https://agentmods.dev/badge/skills/devhimanshulohani/claude-agent-manager/agent.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.00013 | $0.05546 |
| Opus 5 | $0.00006 | $0.02773 |
| Sonnet 5 | $0.00003 | $0.01109 |
| Haiku 4.5 | $0.00001 | $0.00555 |
Grade B, and why
agent scanned grade B 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 5d 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
3. If `taskId` is not null, find the agent's subagent transcript: search for files matching `~/.claude/projects/*/subagents/agent-*.jsonl` using Glob. Look for the transcript that corresponds to the agent's `taskId`. If How it starts
The opening of the file, as written. The whole thing — 283 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Manager v4
Parse $ARGUMENTS and execute the matching command below. Persistent state lives in .claude/agents/registry.json (source of truth). Task system is for live monitoring only.
Pre-flight: Before executing any command, verify the current directory is inside a git repository (git rev-parse --is-inside-work-tree). If not, reject: "Agent manager requires a git repository. Run git init first."
Registry: Read .claude/agents/registry.json — if missing/corrupt, default to { "version": 2, "defaultBranch": null, "agents": [] }. If version is 1 or missing, migrate: add model: null, effort: null, color: null, maxTurns: null to each agent entry, add defaultBranch: null at top level, set version to 2, write immediately. Always mkdir -p .claude/agents before writing.
Atomic writes: Always write registry to .claude/agents/registry.json.tmp first, then rename to .claude/agents/registry.json. This prevents corruption from interrupted writes.
Default branch: Read from registry.defaultBranch. If null (first time using the plugin in this repo), ask the user using AskUserQuestion: "What is your default/base branch? (e.g., main, master, develop)". Store the answer in registry.defaultBranch and write registry. Once set, never ask again. Use {defaultBranch} everywhere in this document instead of a hardcoded branch name.
Agent ID: Generate via echo "$(date +%s)$$.$RANDOM" | shasum | head -c 6 (6-char hex). $$ (PID) + $RANDOM ensures uniqueness even when called multiple times in the same second (e.g., batch).
Age format: <N>m ago / <N>h ago / <N>d ago relative to createdAt.
Prefix match: All ID-based commands support prefix matching. If the prefix is ambiguous (matches multiple agents), reject and show the matching IDs. If no match, show all available IDs.
Templates dir: .claude/agents/templates/ — JSON files with required fields { name, description } and optional { verifyCommand, commitFormat }. verifyCommand overrides Phase 4 auto-detection; commitFormat overrides the default conventional commit format.
User defaults: If env var CLAUDE_PLUGIN_OPTION_DEFAULT_MODEL is set, use it as default model when --model is not specified. If CLAUDE_PLUGIN_OPTION_DEFAULT_EFFORT is set, use it as default effort level. These are configured when the user enables the plugin.
Auto-registry: A SubagentStop hook automatically updates the registry when a worker agent completes. The list command still verifies status as a fallback.
Flag validation: --model must be one of haiku, sonnet, opus. --color must be one of red, blue, green, yellow, purple, orange, pink, cyan. --effort must be one of low, medium, high, max. --max-turns must be a positive integer. Reject invalid values with a clear error message listing valid options.
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.
- 5d ago First seen · 283 lines · 13 tokens per session scan B 1277baa90e3b
agent is a skill published in the GitHub repository devhimanshulohani/claude-agent-manager (6 stars, last pushed 5mo ago), licensed MIT. It adds 13 tokens to every session and 5,546 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
status
Shows MorningStar queue health for a repo -- weekly spend bar, recent runs, success rate, and most recent PR URLs. Use when the user asks any of "morningstar status", "queue health", "how is morningstar doing", "what did morningstar ship this week", "weekly spend so far", "morningstar dashboard", "show recent runs"…
watch
Polls Notion and Jira for pending MorningStar PRDs and dispatches the GitHub Actions executor when work is found. Use when the user says any of "watch the queue", "poll for prds", "check queue", "trigger morningstar now", "is anything queued", "force a queue scan", or "morningstar 24/7 status". Safe to run on a…
dry-run
Previews the MorningStar task plan for a PRD without writing any code or making commits. Fetches the Notion PRD, analyzes the current codebase, and prints the proposed task list with files, acceptance criteria, and an estimated cost. Use when the user says any of "preview", "dry run", "what would this build", "show me…
version
Reports the installed MorningStar plugin version. Use when the user asks "what version of morningstar", "morningstar version", "which morningstar build is this", "is morningstar up to date", or wants to confirm the plugin/agent build before filing a bug.
run
Implements a product requirements document (PRD) end-to-end. Reads the PRD from a Notion page, analyzes the current codebase, generates a task plan, and then implements each task with tests and git commits. Use when the user provides a Notion URL or says any of "implement this PRD", "build from requirements", "run…
citation-audit
Zero-context verification that every bibliographic entry in the paper is real, correctly attributed, and used in a context the cited paper actually supports — catching hallucinated authors, wrong years, fabricated venues, version mismatches, and wrong-context citations. Use when user says "审查引用", "check citations"…