learn

An agent skill by garrytan, from garrytan/gstack. Tags: analysis, documentation, tracking.

What it does

Manage project learnings.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add garrytan/gstack --skill learn

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/garrytan/gstack
cp -r gstack/learn ~/.claude/skills/learn

Safety box score

Grade C, moderate risk: 44 of 100 risk points, rated 2026-10-16. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill manages a local project-learnings store from your home directory, using helper scripts to search, add, prune, export, and log telemetry. The main risks are local file mutation, trusting instruction blocks emitted by a tool, and the fact that much of the behavior lives in external helper scripts that are not visible in this markdown.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	2 (moderate)	It shells out to local helpers like `~/.claude/skills/gstack/bin/gstack-skill-start`, `gstack-learnings-search`, `gstack-learnings-log`, and `gstack-skill-end`, so running the skill executes code on the machine.
Network access	0 (none)	No remote hosts, downloads, or explicit data exfiltration are described; telemetry and artifacts sync are written locally under `$GSTACK_STATE_ROOT`.
Filesystem reach	2 (moderate)	It reads and writes local state in `~/.claude/skills/gstack/`, `.gstack/tmp/qt.txt`, and `$GSTACK_STATE_ROOT/projects/$SLUG/learnings.jsonl`, and prune can rewrite that file.
Secrets & credentials	0 (none)	The skill does not request tokens, passwords, browser sessions, or `.env` files; the only cookie mention is a generic example in prose, not an operation.
Privilege & persistence	1 (low)	It touches home-dot directories and project state, but it does not ask for sudo, system services, global installs, or OS-level configuration changes.
Instruction hijack surface	2 (moderate)	It tells the agent to honor `GSTACK_INSTRUCTION_BEGIN` blocks from tool output and to auto-choose options in spawned/unattended sessions, which widens the prompt surface.
Irreversible actions	2 (moderate)	The prune workflow can remove learning entries after a user choice, and the skill can append/export shared knowledge, but the destructive path is gated by confirmation.
Transparency	2 (moderate)	Behavior is delegated to multiple helper scripts in `~/.claude/skills/gstack/bin/` whose implementations are not shown here, so a reviewer cannot fully audit them from this file.

Before installing

Inspect `~/.claude/skills/gstack/bin/gstack-skill-start` and `gstack-learnings-log` before trusting the workflow.
Confirm `$GSTACK_STATE_ROOT` resolves to a home-directory path you are okay with the agent writing to.
Try `/learn prune` only on a copy of `learnings.jsonl` first if you care about keeping history.
Check the helper scripts for any network calls or writes outside the project state tree.

Rated 2026-10-16 by an automated review of 1 file from the skill's repository.

Source

Repository: garrytan/gstack (all skills from this repository)
Path: learn/SKILL.md
Branch: main
Updated: 2026-10-05

expo-skill-feedback — Submit feedback on an Expo skill—or Expo itself—and control bundled anonymous usage telemetry (off by default / opt-in).
audit-prep-assistant — Prepares codebases for security review using Trail of Bits' checklist.
automate-me — Use for "automate me", "create/update/refresh my -mode skill", "turn/capture my preferences or working style into a skill".
crypto-protocol-diagram — Extracts protocol message flow from source code, RFCs, academic papers, pseudocode, informal prose, ProVerif (.pv).
design-md — Analyze Stitch projects and synthesize a semantic design system into DESIGN.md files
diagramming-code — Generates Mermaid diagrams from Trailmark code graphs.