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 wan-huiyan/agent-traffic-control --skill subagent-driven-branch-ref-froze-stranded-commitsgit clone --depth 1 https://github.com/wan-huiyan/agent-traffic-controlWrote 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/wan-huiyan/agent-traffic-control/subagent-driven-branch-ref-froze-stranded-commits)<a href="https://agentmods.dev/skills/wan-huiyan/agent-traffic-control/subagent-driven-branch-ref-froze-stranded-commits"><img src="https://agentmods.dev/badge/skills/wan-huiyan/agent-traffic-control/subagent-driven-branch-ref-froze-stranded-commits/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/wan-huiyan/agent-traffic-control/subagent-driven-branch-ref-froze-stranded-commits"><img src="https://agentmods.dev/badge/skills/wan-huiyan/agent-traffic-control/subagent-driven-branch-ref-froze-stranded-commits.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00545 | $0.03710 |
| Opus 5 | $0.00272 | $0.01855 |
| Sonnet 5 | $0.00109 | $0.00742 |
| Haiku 4.5 | $0.00055 | $0.00371 |
Grade A, and why
subagent-driven-branch-ref-froze-stranded-commits 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 6d 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.
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 — 313 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Subagent-driven session left commits stranded — branch ref froze while worktree HEAD advanced
Problem
You ran a long superpowers:subagent-driven-development (or similar fresh-
subagent-per-task) session in a worktree. Each subagent did its task, committed
its work, the controlling session ran spec + code-quality review per task, all
green. At the end you pushed the worktree's branch to origin, opened a PR, CI
went green, you squash-merged.
Hours or one session later, somebody runs pytest on freshly-pulled main and
gets a ModuleNotFoundError, a missing-file assertion, or a sql_path missing
error pointing at code that was definitely committed during the session.
git log in your old worktree still shows all the commits at the top of the
branch. But on main, half of them are gone.
The squash commit on main contains only the early portion of your work —
typically the first N tasks before some specific transition point. Tasks
after that point exist as commits in the worktree (git log shows them)
but did not reach origin, did not reach the PR, did not reach the merge.
Context / Trigger Conditions
Strong signal after the merge:
gh pr view <N>body claims a file count (e.g. "55 files changed") that doesn't match the actual squash-commit file count (git diff-tree --no-commit-id --name-only -r <merge-sha> | wc -l).- A test that passed locally during the session now fails on main because a file the test depends on isn't there.
- Receiving session's first
pip installorpytestsurfacesModuleNotFoundError,sql_path missing, or "no such file" referencing work the previous session reported as DONE.
Strong signal during the session (catch it before pushing):
git rev-parse HEAD≠git rev-parse <feature-branch>in the worktree. The branch ref is behind.git reflog show <feature-branch>has a reflog gap — the most recentworktree-...@{0}entry is older than the most recent commit shown bygit log. The commits between branch-ref's frozen point and current HEAD are stranded.git statusin the worktree may or may not reportHEAD detached at <sha>— sometimes the detach happened mid-stream and a later subagent created the symbolic-ref linkage on its own, leaving an inconsistent state.
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.
- 6d ago First seen · 313 lines · 545 tokens per session scan A 5bebe74e790f
subagent-driven-branch-ref-froze-stranded-commits is a skill published in the GitHub repository wan-huiyan/agent-traffic-control (3 stars, last pushed 6d ago), licensed MIT. It adds 545 tokens to every session and 3,710 once invoked, about $0.0027 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 2026-09-05.
Other skills, from other repositories
claude-code-bash-patterns
Claude Code Bash tool patterns with hooks, automation, git workflows. Use for PreToolUse hooks, command chaining, CLI orchestration, custom commands, or encountering bash permissions, command failures, security guards, hook configurations.
git-commit
Turn the working changes into one or more atomic commits with well-written messages. Use whenever the user runs /git-commit or asks to commit their work, wrap up a feature, or "commit what I have.".
misuzu
Respond to PR review comments semi-automatically. Fetches all review threads and comments, splits them into logical units, fixes and commits one unit at a time, then replies and resolves review threads. Use when the user wants to address PR review feedback or respond to review comments.
commit-style
Commit message style guidelines. Use when writing or proposing git commit messages, including direct git commit commands outside the /commit skill.
pre-commit
Fast pre-commit quality gate: scans staged files for secrets, dead code, naming issues, merge conflict markers, and direct-to-main commits. Installs as a git hook via /pre-commit install.
commit
One-shot conventional commit; bundled scripts adaptively gather diff context, draft the message, and stage+commit — pass ask to confirm first or add a hint for the title. Use only when the user asks for a commit.