Borrowing it
Nothing to install: this file belongs to cboone/agent-harness-plugins. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/cboone/agent-harness-plugins/main/.github/instructions/shell.instructions.mdgit clone --depth 1 https://github.com/cboone/agent-harness-pluginsWrote 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/instructions/cboone/agent-harness-plugins/shell)<a href="https://agentmods.dev/instructions/cboone/agent-harness-plugins/shell"><img src="https://agentmods.dev/badge/instructions/cboone/agent-harness-plugins/shell/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/instructions/cboone/agent-harness-plugins/shell"><img src="https://agentmods.dev/badge/instructions/cboone/agent-harness-plugins/shell.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 | $0.01669 | $0.01669 |
| Opus 5.5 | $0.00668 | $0.00668 |
| Sonnet 5.5 | $0.00334 | $0.00334 |
| Haiku 4.5 | $0.00167 | $0.00167 |
Grade A, and why
agent-harness-plugins shell.instructions.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 39 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Shell script review instructions
Bundled scripts here run on macOS and Linux alike, so portability findings are welcome. These specific ones are not defects, and have been checked:
-
git worktree list --porcelain -zis the only form that is parsed. The newline-delimited form splits a worktree path containing a newline across two apparent fields. NUL is the one byte a path cannot hold, so-zplusread -r -d ''is correct for every path a filesystem allows. Aread -d ''loop also needs|| [[ -n "${line}" ]]so a final field carrying no terminator is not discarded. -
flockis deliberately absent. macOS has noflock(1), which is why mutual exclusion here usesmkdir, which is atomic on every filesystem these scripts run on. -
kill -0 <pid>is a liveness test, not a terminate. It is the POSIX interface and keeps its name under the neutral-terminology convention, which exempts names the world already fixed. -
A
"$(< path)"read is always preceded by-fand-rtests. That expansion fails during expansion rather than as a command, so a redirection on the assignment and a trailing||both miss it and Bash exits with its own unprefixed diagnostic.-ralone is insufficient: it is true for a readable directory. A scrut case intests/scrut/repo-tooling.mdenforces this. -
localis declared separately from command substitution (local xthenx="$(cmd)") on purpose:localalways returns 0 and would mask the command's exit code. -
# shellcheck disable=SC2016at file scope is intentional in scripts built aroundjqfilters. The$nametokens inside single-quotedjqprograms arejqvariables bound with--arg, not shell expansions. -
jq'ssubandgsubtake an optional third flags argument.sub/3andgsub/3are builtins alongsidesub/2andgsub/2, sosub("^copilot"; ""; "i")is a case-insensitive substitution, not a call with a surplus operand. Confirm withjq -n 'builtins | map(select(startswith("sub/") or startswith("gsub/")))', which lists all four. The same applies toscan/2,match/2,test/2,capture/2andsplits/2. -
jq -nwithinputsreads every input record, the first included.-nbinds.tonullinstead of the first record; it does not stopjqreading standard input, andinputsconsumes all of it.printf 'a\nb\n' | jq -Rnc '[inputs]'prints["a","b"]. This is the idiom for a single pass that stops early withlabel/break, as innotify'sTURN_JQ. -
gh api --paginatewithout--jqconcatenates REST array pages into one array. A paginated array endpoint yields a single JSON document, so a downstreamtype == "array"check holds at any page count and no flattening step is needed.--jqis what changes this: gh applies the filter to each page separately, so--paginate --jq 'length'over three pages emits three numbers.--slurpis for the array-of-arrays shape and is required only when page boundaries must be preserved, as in a probe that pipes into a standalonejqand takeslastacross all pages. Thegh api --helpline "Each page is a separate JSON array or object" describes the GraphQL and object-returning cases; do not read it as applying to REST array endpoints. -
fetch_thread_indexinresolve-copilot-threadscaps each thread atcomments(first: 100)on purpose. The index is a lookup table, not the audit's source of comments:auditreads every review comment through the paginated REST endpoint and looks each one up in it. A comment beyond the first 100 in a thread is missing from the index and is reported asin no thread, so the cap can mislabel a reason but cannot make the audit report a clean surface. Do not ask for the nested connection to be paginated. -
notify'srun_alerterstops eachalerterwithSIGKILLafterWAIT_LIMITseconds on purpose.SIGTERMandSIGINTmakealerterremove the delivered banner;SIGKILLleaves it in Notification Center. Eachrun_alerterstops only its own childalerter, within 60 seconds at most, so it needs no PID start-time check, lock or shared state. Do not ask for a gentler signal, a lock or a start-time check.
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.
- 3d ago Changed · +2 lines · +94 tokens per session 22c0ba3daf89
- 5d ago Changed · +91 tokens per session 401e5b0a4294
- 6d ago Changed · +7 lines · +368 tokens per session e5f836d6e047
- 8d ago Changed · -1 lines · -141 tokens per session 0265b62ce9bb
- 10d ago Changed · +2 lines · +121 tokens per session 2ba20ffda7c2
- 18d ago Changed · +1 lines · +171 tokens per session 96293d5b55a5
- 21d ago Changed · +1 lines · +177 tokens per session 50eb648d9124
- 26d ago Changed · +1 lines · +113 tokens per session fec5f1c1e782
- 28d ago Changed · +2 lines · +48 tokens per session 6cc1022cf306
- 1mo ago First seen · 24 lines · 627 tokens per session scan A 0eb175424914
agent-harness-plugins shell.instructions.md is an instructions file published in the GitHub repository cboone/agent-harness-plugins (2 stars, last pushed 2d ago), licensed MIT. It adds 1,669 tokens to every session, about $0.0067 per session on Opus 5.5. 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 2026-09-15.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
deepseek-harness AGENTS.md
AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-stable apis and released session data, repository layout, commands and host sandbox failures.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).