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 skills add brunob54/superpowers-orchestrator --skill orchestrating-developmentgit clone --depth 1 https://github.com/brunob54/superpowers-orchestratorWrote 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/brunob54/superpowers-orchestrator/orchestrating-development)<a href="https://agentmods.dev/skills/brunob54/superpowers-orchestrator/orchestrating-development"><img src="https://agentmods.dev/badge/skills/brunob54/superpowers-orchestrator/orchestrating-development.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 12 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Tool Misuse · line 119 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
- high Tool Misuse · line 2167 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
- high YARA Match · line 132 YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).Fix: Remove the malware payload or compromised file entirely. Investigate how it entered the skill and audit all other artifacts for additional indicators of compromise.
- high Prompt Injection · line 162 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
- high Tool Misuse · line 871 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
- high Tool Misuse · line 1028 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
- high System Prompt Leakage · line 1438 Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.Fix: Remove any instructions that reveal, print, or output system prompts or internal rules. System instructions should never be exposed to end users.
- high Prompt Injection · line 2232 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
- medium Agent Snooping · line 206 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
- medium Agent Snooping · line 211 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
- medium Excessive Agency · line 343 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- low Privilege Escalation · line 197 Skill requests more permissions than appear necessary for its stated functionality. Review if elevated access is justified.Fix: Request only the minimum permissions required. Document why each permission is needed. Remove broad permissions like '*' or 'all'.
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 5.1 | $0.00120 | $0.35461 |
| Opus 5 | $0.00060 | $0.17730 |
| Sonnet 5 | $0.00024 | $0.07092 |
| Haiku 4.5 | $0.00012 | $0.03546 |
Grade A, and why
orchestrating-development scanned grade A with 0 findings 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 today.
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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 2,258 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Orchestrating Development
Drive an approved spec through plan → plan review → batched implementation
→ whole-branch code review with no user interaction between Phase 0 and
completion. You are a thin sequencer: every phase and every batch runs in a
fresh controller subagent; all state moves through files. Never read plan
bodies, diffs, reviewer reports, or fix reports yourself. Two documented
exceptions: Phase 0 step 4's prior-art intake check reads
the spec body once, before any controller dispatch; and the
classification read of ## In-run rulings ("What may be read"), which is
bounded to the list stated there — nothing else.
Required Start
Announce: I'm using orchestrating-development to run this pipeline.
Platform check: this skill requires the Agent tool (with nested
dispatch). On platforms without it, refuse with one line —
orchestrating-development requires subagent dispatch (Agent tool), which this platform lacks — and stop.
Controller Dispatch Rules (apply to every phase)
- Dispatch via the Agent tool,
general-purposetype. Model: inherit the session model with a sonnet floor (a haiku-tier or unrecognized session model dispatches onsonnet). Nested workers inside a batch follow SDD's Model Selection table, chosen by the batch controller. - Blocking dispatch: pass the
namestated by the phase's dispatch text (Phases 1–4) (name: "orch-…"). A named controller's own subagent calls block and return each child's final message inline; an unnamed controller's return at once, and the child's completion notice is delivered to the main session, not to the controller — which then ends its turn "waiting" and stalls the run (claude-code #75043; measured 2026-08-30). The property is what matters, not the parameter: on a platform whose dispatch tool has noname(Copilot CLI:task/agent), the controller obtains it from the "Waiting on a subagent" rule in its prompt — foreground dispatch, never a background mode. Controllers cannot name their own children (the roster is flat); nested workers stay unnamed. - Prompt files and the pointer. A controller's instructions are a
file, not the
promptfield. Once per session, read as once per orchestration invocation — a fresh run, or a Resume, including a Resume inside the session that stopped — before that invocation's first fill — Phase 0 step 9 in a fresh run; on a resume, the Resume step that first fills a prompt — runmktemp -das its own Bash command, with no argument (never a template and never a path inside the repository), and copy the literal path it prints — written<PROMPT_DIR>in this file — into every later command, Write call and pointer. A printed path under the repository root is treated as amktemp -dfailure. On Git Bash (Windows) — whenuname -sprints a name beginning withMINGWorMSYS— first convert that path once withcygpath -m "<printed path>"as its own command and use the converted path as<PROMPT_DIR>; on every other platform the printed path is used as is. A shell variable set in one tool call does not exist in the next: the path is always spelled out in full, never held in a variable of any name. The path never appears instate.md, the orchestration log, a ruling record, a Case or a commit message. When the literal path is no longer in your context — after a context compaction, typically — do not guess it, do not search the temporary location for it and do not reuse a path from any file: runmktemp -dagain and continue in the new directory, where the counter<k>below restarts at 1. That is not a failure of the mechanism: every file is named by the counter, and a re-fill from the same values produces the same content, so an identical retry issued after the path was lost re-fills the prompt under<k>= 1 there — writing the value file again there first when the fill takes one — and is still the identical dispatch of the retry rule. Every prompt is written once into<PROMPT_DIR>by the fill script,<base>/../multi-code-review/scripts/fill-prompt.js(<base>is this skill's base directory), from the phase's template and the values the phase's fill command lists; you never read a template and never copy a prompt body into an Agent call. File names, all in<PROMPT_DIR>, where<k>counts the fills run into the current prompt directory from 1 and is never reused inside it once a pointer to a file of that fill has been dispatched, and<label>names the dispatch:
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- today Changed · +1,740 lines 32d14aba4454
- 7d ago First seen · 518 lines · 120 tokens per session scan A c79192db4820
orchestrating-development is a skill published in the GitHub repository brunob54/superpowers-orchestrator (3 stars, last pushed yesterday), licensed MIT. It adds 120 tokens to every session and 35,461 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
github-pr-review
Handles PR review comments and feedback resolution. Use when user wants to resolve PR comments, handle review feedback, fix review comments, address PR review, check review status, respond to reviewer, verify PR readiness, review PR comments, analyze review feedback, evaluate PR comments, assess review suggestions, or…
github-pr-creation
Creates GitHub Pull Requests with automated validation and task tracking. Use when user wants to create PR, open pull request, submit for review, or check if ready for PR. Analyzes commits, validates task completion, generates Conventional Commits title and description, suggests labels. NOTE - for merging existing…
assess
Issue triage and status assessment - analyze current state, detect health signals, and recommend next action with full workflow plan.
fullsolve
Complete issue resolution with integrated quality loops - spec → exec → test → qa with auto-fix iterations.
loop
Quality loop - Parse test/QA findings and iterate until quality gates pass.
merger
Multi-issue integration and merge skill - handles post-QA integration of completed worktrees.