AgentStack
SKILL verified MIT Self-run

Altman Mode Install

skill-tarcroi-altman-mode-altman-mode-install · by Tarcroi

>

No reviews yet
0 installs
2 views
0.0% view→install

Install

$ agentstack add skill-tarcroi-altman-mode-altman-mode-install

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access Used
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Altman Mode Install? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Altman-mode Install

Install or update the altman-mode CLAUDE.md guidelines in the user's current project — non-destructively and idempotently.

This skill ships with the altman-mode Claude Code plugin. When invoked, it ensures the project's CLAUDE.md contains the altman-mode rules without touching anything the user already wrote.

When to use this skill

  • The user asks to install, add, set up, configure, or apply altman-mode in the current project.
  • The user wants the altman-mode CLAUDE.md guidelines dropped into this folder.
  • The user just installed the altman-mode plugin and wants to enable it for the current project.

When NOT to use this skill

  • The user wants to score a startup idea → use altman-grader.
  • The user asks generic questions about CLAUDE.md without asking to install it.
  • The user wants to remove altman-mode → different operation.
  • Code reviews, refactors, or anything unrelated to project setup.

Expected inputs

No structured input required. The skill operates on the user's current working directory.

If the user mentions custom rules or pre-existing content ("I already have a CLAUDE.md, don't break it"), still follow the procedure below — preserving existing content is the default behavior.

Procedure

Step 1 — Fetch the source CLAUDE.md

Always fetch the latest content from the canonical URL:

https://raw.githubusercontent.com/Tarcroi/altman-mode/main/CLAUDE.md

Use Bash with curl. Always include --ssl-no-revoke:

curl -fsSL --ssl-no-revoke https://raw.githubusercontent.com/Tarcroi/altman-mode/main/CLAUDE.md

The --ssl-no-revoke flag is required on Windows: the schannel TLS backend often fails with CRYPT_E_NO_REVOCATION_CHECK when it cannot reach the certificate's revocation server. The flag is silently ignored on macOS and Linux, so it is safe everywhere — keep it in the command unconditionally.

If curl fails for any other reason (no network, 404, empty response), abort and tell the user the source could not be fetched. Do not write anything.

Sanity-check the fetched content: if it is fewer than 100 lines, treat as corrupted and abort.

Step 2 — Detect existing CLAUDE.md

Check whether ./CLAUDE.md exists in the user's CWD:

test -f ./CLAUDE.md && echo exists || echo missing

Step 3 — Branch on three cases

Case A — No existing CLAUDE.md

Write the fetched source content directly to ./CLAUDE.md. Do not add BEGIN/END markers in this case — the entire file is the altman-mode CLAUDE.md and there is nothing to delimit it from.

Confirm to the user:

> Created CLAUDE.md (X lines) with the altman-mode guidelines.

Case B — Existing CLAUDE.md, NO markers present

Read the existing file. Append the following block at the very end, preceded by a single blank line:


## Altman-mode guidelines (from altman-mode plugin)

[fetched source content, with its leading "# CLAUDE.md" heading stripped if present, since we already have a heading on the line above]

Use the Edit tool or read-then-Write the full file. Either way the user's original content above must be byte-identical after the operation.

Confirm to the user:

> Appended altman-mode guidelines below your existing CLAUDE.md (X lines added). Re-run this skill anytime to update the altman-mode block in place.

Case C — Existing CLAUDE.md, markers ALREADY present

Locate the lines containing ` and . Replace **only** the content strictly between them with the freshly fetched source (with its leading # CLAUDE.md heading stripped, and the same ## Altman-mode guidelines (from altman-mode plugin)` heading kept on the first line inside the block). The markers themselves stay in place.

Content above the BEGIN marker and below the END marker must be byte-identical to before.

Confirm to the user:

> Updated the altman-mode block in your CLAUDE.md (X lines, replacing Y previous lines).

Step 4 — Confirmation summary

End with a one-line factual summary: created / appended / updated, plus the line count.

Do not narrate progress. Do not add emoji unless the user asked for them. Do not include the full content in the response — the user can read the file.

Output format

The skill produces:

  1. A modified or newly created ./CLAUDE.md in the user's CWD.
  2. A short factual confirmation message.

Nothing else.

Quality criteria

A good run:

  • Never modifies user content above the altman-mode block.
  • Is byte-identical idempotent on re-run when the source has not changed.
  • Updates in place (no duplicate blocks) when re-run with a newer source.
  • Tells the user exactly what changed and why it's safe.

A bad run:

  • Overwrites the entire existing CLAUDE.md.
  • Adds duplicate altman-mode blocks on re-run.
  • Leaves stale or partial content between the markers.
  • Modifies the user's prose, formatting, or whitespace outside the markers.

Error handling and edge cases

Network failure on curl. Abort. Tell the user: "Could not fetch the altman-mode source from GitHub. Check your connection and re-run."

Fetched content is suspiciously small (< 100 lines) or empty. Likely a fetch error or 404 page. Abort. Tell the user the source looks corrupted, do not write anything.

Mismatched markers (only BEGIN, only END, or multiple BEGIN/END pairs). Treat as malformed. Tell the user the markers are mismatched, do not modify the file, and ask whether to repair the malformed block or leave it alone.

Read-only filesystem or permission error. Report directly with the OS error. Do not retry silently.

User explicitly asks for a different file path (e.g., "install it in ~/.claude/CLAUDE.md"). Confirm the path with the user, then apply the same three-case logic to that path.

Examples

Example 1 — Fresh project, no CLAUDE.md

Input: "Install altman-mode in this project."

Result: Creates ./CLAUDE.md with the fetched content. Reports the line count.

Example 2 — Project with existing CLAUDE.md (no markers)

Input: "Add altman-mode guidelines here, but don't break my existing rules."

Result: Reads existing CLAUDE.md, appends the altman-mode section wrapped in BEGIN/END markers below the existing content. Reports lines added.

Example 3 — Re-run after a few weeks (markers already present)

Input: "Update altman-mode in this project."

Result: Replaces only the content between the markers with the latest source. Reports lines replaced. No duplicates.

Example 4 — User asks to score an idea (anti-trigger)

Input: "Score my startup idea."

Result: This skill does NOT fire. The altman-grader skill handles scoring.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.