shell

Applies working-file placement and shell command discipline for personas running Bash and gh commands.

Updated Jun 4, 2026
One-click install
npx skills add https://github.com/tedeuxx/tadeumendonca-skills --skill shell-tedeuxx
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: shell
Source: https://github.com/tedeuxx/tadeumendonca-skills/tree/main/skills/shell
Command: npx skills add https://github.com/tedeuxx/tadeumendonca-skills --skill shell-tedeuxx

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Personas writing scratch files and running shell commands kept restating the same rules in every agent brief, and unsafe patterns like stdout redirects, command substitution, and inline multi-line PR bodies caused permission prompts and corrupted content. ## Core Features & Use Cases - Working-file discipline: Routes every scratch file (commit messages, PR/issue bodies, drafts) to the session scratchpad via the Write tool, never a shell redirect. - Shell command rules: Documents what the permission matcher actually stops — $(...)/backticks, VAR=x env prefixes, file-creating redirects — and prescribes git -C, npm --prefix, and gh <subcommand> --repo flag positioning. - --body-file mandate: Requires posting any multi-line or backtick-bearing PR/issue body from a file with --body-file so shell interpolation cannot strip backticks and $. - Use Case: A reviewer persona needs to post a multi-line merge verdict with code snippets; it writes the verdict to the session scratchpad with Write, then runs gh pr comment --body-file <path> so the markdown survives intact. ## Quick Start Preload this skill in any persona that writes files or runs shell commands, then follow its rules when composing Bash calls and gh PR/issue bodies.

Frequently Asked Questions about shell

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I post a multi-line GitHub PR comment with gh without losing backticks?

Write the body to a file in the session scratchpad using the Write tool, then post it with `gh pr comment --body-file <path>`. Inline `--body` strings lose backticks and `$` to shell interpolation, and `--body` is not an acceptable fallback.

Why does my gh command trigger a permission prompt when the subcommand is allowlisted?

The permission matcher reads a command prefix, so placing the flag before the subcommand (`gh -R owner/repo issue view`) changes the prefix to `gh -R` and matches no allowlist entry. Put the flag after the subcommand: `gh issue view --repo owner/repo`.

Can I use shell redirects to write scratch files in a Claude Code session?

No. A redirect that creates a file (`command > path`) is denied by the permission guard regardless of destination. Use the Write tool for content you compose, or run the command without the redirect and Write its returned output if it must persist.

Does chaining commands with && or ; trigger permission prompts?

No. The matcher evaluates compositions element-wise: a chain of allowlisted commands executes without prompting, and a chain containing a denied element is caught by that element. One atomic command per call is still preferred for transcript legibility, not permission necessity.

Why does `FOO=1 command` require approval when the bare command is allowlisted?

An env-var prefix defeats the allowlist entry, including in post-statement position after `;`. Prefer the repo's own scripts (e.g., `npm --prefix <app> run <script>`) over inline env-prefixed commands. Note that `=` in argument position, like `make FOO=1 target`, is not an env prefix.

When should this shell discipline skill not be used?

Do not use it for the loop's state machine logic (covered by agents-configuration) or for what a merge verdict must contain (covered by quality-gates). It only governs working-file placement and shell command composition.