Install
$ agentstack add skill-saikaaco-saika-hermes-skills-hermes-workspace-hygiene ✓ 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
Hermes Workspace Hygiene
Overview
Keep Hermes-generated artifacts predictable, scoped, and recoverable:
- temporary/intermediate files go under
/scratch/; - generated reports intended for reading or sending go under
/reports/; - explicit user-requested exports go under
/exports/; - project source remains in its project root;
- existing user files are never moved or deleted without explicit approval.
Resolve the active workspace from project instructions or the current working directory. If no safe workspace root is clear, ask before writing rather than defaulting to a personal home directory.
When to use
Use when:
- a task will create scripts, intermediate JSON/CSV/HTML, screenshots, downloads, report assets, or exports;
- the user asks to audit or clean generated files;
- generated artifacts appear in a high-level personal directory;
- a recurring job needs a predictable artifact and cleanup lifecycle;
- a report or export needs the correct destination.
Do not use for normal source edits inside an existing project, read-only inspection, or durable knowledge and policy files that already have an authoritative location.
Canonical workspace shape
/
scratch/ temporary and intermediate artifacts
reports/ generated reports intended for reading or delivery
exports/ explicit user-requested exports
Prefer topic/profile subfolders:
scratch//
reports//
exports//
Do not create one-off files at a home-directory root. Do not move project-specific source into a generic scratch tree.
Operating rules
- Before writing: choose the narrowest appropriate workspace subfolder.
- Before cleanup: inspect the exact candidates and produce a manifest.
- Before moving or deleting existing files: obtain explicit approval for that manifest.
- Prefer reversible moves over deletion unless deletion is explicitly requested.
- Keep rollback possible: record source, destination, size, reason, and timestamp.
- Fail closed: missing, changed, moved, symlinked, or ambiguous targets require a fresh audit and approval.
- Do not expand scope: newly discovered or newly stale paths do not inherit an older approval.
- Report honestly: distinguish proposed, moved, deleted, skipped, and failed items.
Cleanup manifest
Use a machine-readable manifest such as:
{
"manifest_id": "",
"created_at": "",
"operation": "move | delete",
"items": [
{
"from": "",
"to": "",
"size_bytes": 12345,
"reason": "generated temporary artifact",
"identity_check": ""
}
]
}
The user-facing summary must include item count, total size, operation type, destination, rollback method, and the approval requirement.
An approval binds only to the exact manifest and targets shown to the user. Re-running an audit creates a new scope and requires new approval.
Audit procedure
- Define the allowed root and artifact classes.
- Inspect candidates without mutation.
- Exclude normal user folders, source trees, hidden configuration, and ambiguous files unless explicitly in scope.
- Resolve symlinks and record stable target evidence.
- Produce the manifest and concise decision summary.
- Wait for approval.
- Re-verify every target against the approved manifest immediately before execution.
- Execute only the approved operation and paths.
- Verify resulting locations or absence.
- Report completed, skipped, changed, and failed items plus recovery instructions.
Burnable brief lifecycle
Temporary visual briefs belong under scratch/visual-briefs/. A brief may carry an expiry timestamp, but expiry only makes it eligible for a cleanup proposal; it never authorizes automatic deletion. Durable evidence must remain elsewhere so the brief can expire safely.
Common pitfalls
- Home-root clutter. Tool defaults are not a safe artifact policy.
- Audit-as-approval. Discovering a candidate does not authorize mutation.
- Scope drift. A second audit cannot be executed under the first manifest.
- Path-only identity. Re-check symlinks and changed targets before execution.
- Delete-first cleanup. Prefer recoverable moves when they meet the goal.
- Ambiguous ownership. Ask rather than classifying personal files as generated clutter.
- Artifact/source confusion. Scratch space is not a project source root.
Verification checklist
- [ ] Workspace root and artifact class were resolved before writing.
- [ ] Temporary files, reports, and exports use their intended subfolders.
- [ ] No one-off artifact was written to a personal home-directory root.
- [ ] Cleanup remained read-only until an exact manifest was approved.
- [ ] The execution scope exactly matched the approved manifest.
- [ ] Targets were re-verified immediately before mutation.
- [ ] Every move has a rollback path; every deletion was explicitly approved.
- [ ] Results distinguish completed, skipped, changed, and failed items.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: SaikaAco
- Source: SaikaAco/saika-hermes-skills
- License: MIT
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.