memU— agent skill

Personal memory across agents

by NevaMind-AI·Agent Skill·github.com/NevaMind-AI/memU

Is memU safe to install?

SaferSkills independently audited memU (Agent Skill) and scored it 71/100 (yellow). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 7 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.

Score
71/100
●●●●●●●○○○
Re-scan~30s
Latest scan
ScannedJun 22, 2026 · 134d ago
Scans run0 over 90 days
Detectors55 checks · 5 categories
Findings7 warnings · 0 high
EngineSaferSkills 755a9a5
View methodology →
SaferSkills installs
This week0
This month0
All time0
CategoryWeightCategory scoreContribution
Securityprompt, exec, net, exfil, eval
35%
16
5.6 pts
Supply chainhash, typosquat, maintainer, lockfile
20%
100
20.0 pts
Maintenancestaleness, pinning, CI
15%
100
15.0 pts
TransparencySKILL.md, perms, README
15%
100
15.0 pts
Communityinstalls, verify, response
15%
100
15.0 pts

Findings & checks · 7 flagged

Securityscore 16 · 7 findings
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · src/memu/hosts/codex/INSTALL.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptsrc/memu/hosts/codex/INSTALL.md· markdown
176IP, a WSL or VM host address) still needs the `NO_PROXY` exemption on every
177release, with that address in the list. Either way this is a mechanical
178requirement with exactly one right answer — apply it and move on; do not ask
179the user.
180 
Occurrences
1 occurrence · at L178
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25612211d7a427c799a…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · src/memu/hosts/cursor/INSTALL.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptsrc/memu/hosts/cursor/INSTALL.md· markdown
158`NO_PROXY` exemption on every release, with that address in the list. This is
159a mechanical requirement with exactly one right answer — apply it and move on;
160do not ask the user.
161 
162**The existing config fails doctor? Repair the connection, never the identity.**
Occurrences
1 occurrence · at L160
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25612211d7a427c799a…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · src/memu/hosts/hermes/INSTALL.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptsrc/memu/hosts/hermes/INSTALL.md· markdown
159`NO_PROXY` exemption on every release, with that address in the list. This is
160a mechanical requirement with exactly one right answer — apply it and move on;
161do not ask the user.
162 
163**The existing config fails doctor? Repair the connection, never the identity.**
Occurrences
1 occurrence · at L161
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25612211d7a427c799a…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · src/memu/hosts/generic/INSTALL.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptsrc/memu/hosts/generic/INSTALL.md· markdown
175`NO_PROXY` exemption on every release, with that address in the list. This is
176a mechanical requirement with exactly one right answer — apply it and move on;
177do not ask the user.
178 
179**The existing config fails doctor? Repair the connection, never the identity.**
Occurrences
1 occurrence · at L177
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25612211d7a427c799a…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · src/memu/hosts/openclaw/INSTALL.md
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptsrc/memu/hosts/openclaw/INSTALL.md· markdown
160`NO_PROXY` exemption on every release, with that address in the list. This is
161a mechanical requirement with exactly one right answer — apply it and move on;
162do not ask the user.
163 
164**The existing config fails doctor? Repair the connection, never the identity.**
Occurrences
1 occurrence · at L162
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25612211d7a427c799a…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · src/memu/hosts/claude_code/INSTALL.md×2
MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.
Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted
excerptsrc/memu/hosts/claude_code/INSTALL.md· markdown
207`NO_PROXY` exemption on every release, with that address in the list. This is
208a mechanical requirement with exactly one right answer — apply it and move on;
209do not ask the user.
210 
211**The existing config fails doctor? Repair the connection, never the identity.**
Occurrences
2 occurrences · first at L209, also L253
Show all 2 locations
Line
File
L209
src/memu/hosts/claude_code/INSTALL.md
L253
src/memu/hosts/claude_code/INSTALL.md
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25612211d7a427c799a…rubric 365aacaView on GitHub
Supply chainscore 100 · 0 findings
✓—All supply chain checks passedNo findings in this category for the latest scan.pass
Maintenancescore 100 · 0 findings
✓—All maintenance checks passedNo findings in this category for the latest scan.pass
Transparencyscore 100 · 0 findings
✓—All transparency checks passedNo findings in this category for the latest scan.pass
Communityscore 100 · 0 findings
✓—All community checks passedNo findings in this category for the latest scan.pass
Vendor response · right of reply
Are you the maintainer? Submit a response →

Audit the pieces. Scan the whole. Decide.

~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.