Install
$ agentstack add skill-useaitechdad-agent-skills-brief-delegate ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo 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 No
- ✓ 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.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
brief-delegate — focused task brief for a sub-session
You are writing a brief that delegates one task to a fresh Claude Code session. The sub-session must be able to execute the task from this brief alone — without the parent's context.
Where to save
~/.claude/projects//briefs/brief-.md
Compute the paths yourself — no shell needed for the slug:
- `
= current working directory with every/replaced by-(e.g. cwd/Users/you/code/myapp→ slug-Users-you-code-myapp`). You already know the cwd from your environment. - briefs dir =
/.claude/projects//briefs(absolute, your real home path).
You do need one shell call — the timestamp, since Claude has no clock:
date +%Y-%m-%d-%H%M%S # local time — keeps the filename stamp consistent with the log stamp
The filename is brief-.md. Write it with the Write tool (it creates the briefs dir if missing) — don't shell out to create the file.
Before writing
If the user hasn't already told you, ask once for:
- The single task to delegate (one sentence)
- Success criteria (one or two bullets)
Don't proceed with a vague task — the sub-session will spin.
What to write
# Brief —
**Flavor**: delegate
**Created**:
**Project**:
**Read first**: `CLAUDE.md`, then [the SPECIFIC docs the sub-session needs for THIS task — scan the repo and name them (spec, design doc, the module's README). A bare `CLAUDE.md` is not enough; the sub-session has none of the parent's context.]
## Task
## Context
## Steps
1.
2.
3.
## Success criteria
-
## Constraints
-
## Report back
Commit at logical checkpoints as you go (if this is a git repo) — those SHAs are how the parent reviews exactly what changed. When done, write a new brief using the `brief-report` skill (`/brief-report`). It will save to `~/.claude/projects//briefs/`. Set its **Responds to:** to THIS brief (`brief-.md`). Include:
- What changed (files + commit SHAs)
- What worked / what didn't
- Any new blockers or surprises
When you print this brief's path to the user, the path you give them IS the file the sub-session reads — so the sub-session always knows which brief to cite as Responds to.
After writing
- Append the log entry via the shared helper (owns the format + trim + file pruning — don't hand-format the line). Pass the absolute briefs dir and the file you just wrote:
``bash ~/.claude/skills/_brief-shared/append-log.sh "" delegate "" "" ``
- Print the file path.
- Tell the user verbatim: "Open a new Claude Code session in this project, then say:
Read and execute it." — they could also say/brief-resumeand pick this one from the list.
What NOT to include
- Don't include exploratory ideas — the sub-session does one thing.
- Don't pad with "consider also..." or "you might want to..." — every line is an instruction.
- Don't include reasoning traces or alternatives that were rejected.
- Don't leave success criteria vague ("looks good", "feels right" — never).
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: useaitechdad
- Source: useaitechdad/agent-skills
- License: Apache-2.0
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.