Install verification: runtime verified (L5) · tested on dsh 0.2.1-rc.2 · last checked Install verification details →
What it does
Lossless visual memory for DSH agents: store an image once, address it by sha256, read exact pixels, compare past frames, recall by text/tag/turn — no second vision model in the loop.
Installation
dsh plugin --profile web add github:xiciyang/dsh-visual-memory#v0.1.1Install method: npm · runtime verified (L5)
Verification & Compatibility
Grouped by plugin version: each card lists the dsh versions actually tested against that artifact version and their install verdicts (L5 wins). Untested combinations are omitted — absence means untested, never assumed compatible. L1–L3 are a plugin-level static archive (independent of the dsh version — the same verdict on every dsh row): they sit in each row's level grid next to L4/L5, marked "static", while their evidence is stored once per plugin inside that row's "checks & evidence" instead of being duplicated per version.
External credentials (plugin-level): none detected
○ The author declares no dsh version requirement — install is not blocked by version.
| plugin \ dsh | dsh 0.2.1-rc.2 current | dsh 0.2.1-alpha.4 |
|---|---|---|
| 0.1.6 | · | · |
| v0.1.1 | ✓ | · |
0.1.6npm2026-10-21
dsh-visual-memory
No runtime install test for this version (static scan only) — compatibility unknown, never assumed.
v0.1.1github✓ 1 runtime-verified2026-10-19
https://github.com/xiciyang/dsh-visual-memory
✓dsh 0.2.1-rc.2current2026-10-19L5 · runtime verified
- • Node.js: v24.14.0
- • DSH: 0.2.1-rc.2
- • Plugin artifact: github · https://github.com/xiciyang/dsh-visual-memory#v0.1.1
ID: dsh-visual-memory@v0.1.1@dsh0.2.1-rc.2 · profile: l4-sandbox
Issued: 2026-10-19 · expires: 2026-10-26
View checks & evidence
↳ carried from earlier record dsh-visual-memory@0.1.6@dsh0.2.1-rc.2 (l3-fullscan)
↳ carried from earlier record dsh-visual-memory@0.1.6@dsh0.2.1-rc.2 (l3-fullscan)
↳ carried from earlier record dsh-visual-memory@0.1.6@dsh0.2.1-rc.2 (l3-fullscan)
active in loader tree (150 entries)
L1–L3 are plugin-level static conclusions (baseline, identical across dsh versions): evidence is stored once per plugin.
Only the latest 3 published versions are shown; source-snapshot (commit hash) groups, unstated groups and older versions are omitted — the full archive stays in the data layer (snapshot/unstated groups appear only when a plugin has no published version).
Author badges
Embed the official badges in your README to showcase security & install-test status:
`[](https://www.dsh.so/artifact/dsh-visual-memory/)``[](https://www.dsh.so/artifact/dsh-visual-memory/)`Security Report
Two automated layers, neither a manual review: a static scan (per-plugin rule checks) and a model-driven dynamic security audit that covers only the subset static-flagged at warn or above and never executes plugin code.
Static low, dynamic medium — stricter wins → overall medium.
v0.1.6dsh manifest0.1.6npm2026-10-210.1.6same versionclean · npmVet is an AST static scan (npm artifact or git source).
show static findings (8)
- warningshellExec · shell
lib/index.mjs:181const child = execFile('python3', [IMAGING], { maxBuffer: 32 * 1024 * 1024 }, (error, stdout, stderr) => { - warningnetIp · network
lib/index.mjs:1132const url = new URL(req?.url ?? '/', 'http://127.0.0.1') - warningfsWrite · files
lib/index.mjs:126await fs.writeFile(file, JSON.stringify(next, null, 2), 'utf8') - warningfsWrite · files
lib/index.mjs:228await fs.appendFile(path.join(root, INDEX_FILE), JSON.stringify(record) + '\n', 'utf8') - warningshellExec · shell
verify.mjs:43const stdout = execFileSync('python3', [IMAGING], { input: JSON.stringify(request), maxBuffer: 32 * 1024 * 1024 }) - warningfsWrite · files
verify.mjs:161fs.writeFileSync(secret, 'TOP-SECRET-DO-NOT-SERVE') - warningfsWrite · files
verify.mjs:188fs.writeFileSync(path.join(store, 'index.jsonl'), JSON.stringify(record) + '\n', 'utf8') - warningfsWrite · files
verify.mjs:189fs.writeFileSync(path.join(dshHome, 'visual-memory-roots.json'), JSON.stringify([{ root: store }]), 'utf8')
show 3 audit findings
- mediumRaw route serves index-supplied originalHostPath without a containment check
lib/index.mjs:157Variant paths (`v=`) are realpath-fenced inside <root>/artifacts, but the original branch returns `record.originalHostPath` straight from index.jsonl, which lives at <session cwd>/.dsh-visual-memory/index.jsonl — a directory the agent can write with its fs tools. A poisoned record (id + originalHostPath=/etc/passwd or ~/.ssh/id_rsa, format='svg+xml') makes GET /_dsh/visual-memory/raw?id=… return those bytes with the record-chosen Content-Type. The request still has to pass ctx.connection.requestRejection (fail-closed Host/Origin fence + browser auth), so this is not remotely unauthenticated, b
- lowGallery store-root allow-list falls back to a workspace-writable registry file
lib/index.mjs:111When DSH_HOME is unset, visual-memory-roots.json is read from and written to process.cwd(), so a party able to write that directory (the agent itself) can register an arbitrary directory as a served store root; the route then trusts that directory's index.jsonl and the paths it lists, widening the originalHostPath issue. It also differs from the ~/.dsh fallback used by verify.mjs.
- lowGallery interpolates unescaped numeric and ROI fields from the index
lib/index.mjs:510String fields (id, note, text, tags, paths) are correctly passed through escapeHtml, but width, height, bytes and the roi coordinates (line 497) are concatenated raw. A crafted index record with e.g. width = '"><img src=x onerror=…>' yields script execution in the DSH web origin when an authenticated user opens the gallery. Precondition is the same agent-writable index.jsonl.
Only lib/index.mjs runs (plus a python3 child); the browser gallery/raw route trusts the agent-writable index for originalHostPath and numeric HTML fields — a guarded file-read/XSS chain.
Rule: the dynamic audit can only raise risk, never lower it (raise-only, never an exemption); a wrong raise is withdrawn by human review and the tier falls back to the static verdict. Overall verdict = the stricter of the two layers.
Disclaimer: the static scan and the dynamic security audit are both automated analysis, not a security guarantee. Always review what you install.
Activity
Last commit 2026-10-19 · activity: Active
• Repo created: 2026-10-17
• Stars: ★ 2 · Forks: 0
• Health: Active — committed within last 30 days
Source
Author README
View on GitHub ↗The author's original README from the repository — fetched live, unedited (the English page loads the default README).
