Skills/Glama/mazarbul

mazarbul@r1

CAUTION· install only

Publisher claims: "Local-first, git-synced memory and system of records for AI agents: a signed append-only event log projected into graph, vector and full-text indexes, with hybrid recall and provenance on every fact." Observed in sandbox: calls api.github.com only, touches nothing outside the project, writes no memory.

Registry GlamaPublisher mo-cmyk · publishing for 3 daysInstalls 0Last detonated 3d ago
What would fix thisWatchDispute verdict

Coverage

Install scripts ran; no agent task scenario for this bundle.
Tested
install scripts, with decoy credentials in place · second launch with the clock 45 days ahead · every file read, file written and host contacted by what ran
Not tested
use of the skill by an agent

What this means for the label: Caution here covers what ran. The server's tools were not exercised with a real credential, so nothing about their behavior is asserted; the credential summary below says what to do with that.

Before you give it your token

It will ask for
ANTHROPIC_API_KEYGITHUB_TOKENMAZ_INSTALL_TOKENConfiguration, not credentials: FILE_PATH, MAZ_NO_MODIFY_PATH, MODIFY_PATH, SEARCH_PATH
Where it talks to
no vendor host named in the codeAlso githubusercontent.com, ollama.com — not the publisher's domain; check what for before connecting.
What the honeypot saw
no credential files read · reached only api.github.com
What we could not test
use of the skill by an agent

Connect it with a scoped credential, under the guard, and read the third-party hosts first. The code names githubusercontent.com, ollama.com as well as its own vendor. The guard injects ANTHROPIC_API_KEY, GITHUB_TOKEN, MAZ_INSTALL_TOKEN only into requests to the host you bind (and its subdomains) and refuses the rest.

$vet-skills run --net declared -- <the server's command>

Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.

Is mazarbul@r1 safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios; install scripts ran; no agent task scenario for this bundle) on 2026-11-11 and scored it 62/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: False input termination marker; Collapsed <details> block carrying instructions; Pipe-to-shell pattern. Summary: calls api.github.com only, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 1 s · 2 scenarios · 0 deviations
t+CategoryObserved action
0.2snetworkCONNECT tcp://api.github.com:443
0.2sprocessexec /usr/bin/uname
0.2sprocessexec /usr/bin/curl
0.2sprocessexited 1: install.sh: could not query the stable release (token valid? repo access?)
0.2sprocessscript backend runs setup only
0.2sprocessscript backend runs setup only
0.4sprocesssecond launch, clock +45 days, CI unset
0.6snetworkCONNECT tcp://api.github.com:443
0.6sprocessexec /usr/bin/uname
0.6sprocessexec /usr/bin/curl
0.6sprocessexited 1: install.sh: could not query the stable release (token valid? repo access?)

Static findings

CRITICAL
False input termination marker.claude/skills/babysit-abort/SKILL.md:6 · False input termination marker — attempts to end user input section---Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
HIGH
Collapsed <details> block carrying instructionsREADME.md:58 · Collapsed <details> block carrying instructions<summary><b>Build from source</b></summary> ```bash # Prerequisites: mise (https://mise.jdx.dev), git. git clone https://github.com/mo-cmyk/mazarbul.git cd mazaFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
CRITICAL
Pipe-to-shell patternscripts/install.sh:5 · Pipe-to-shell pattern — downloads and immediately executes remote code# curl -fsSL https://raw.githubusercontent.com/mo-cmyk/mazarbul/main/scripts/install.sh | shFix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST06
HIGH
Direct script execution from GitHub raw URLscripts/install.sh:5 · Direct script execution from GitHub raw URL — unverified code source# curl -fsSL https://raw.githubusercontent.com/mo-cmyk/mazarbul/main/scripts/install.sh | shFix: Pin versions and sources; vendor anything fetched at install time.
AST06
HIGH
Access to sensitive file paths.claude/skills/devstack/SKILL.md:59 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files- **Port collision between worktrees**, devstack picks a port window per worktree; if collision still happens, check `.env.local`.Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
HIGH
Access to sensitive file paths.claude/hooks/session-namer.sh:106 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files# .env.local if available; fall back to basename of root.Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
CRITICAL
Pipe-to-shell patternscripts/install.sh:348 · Pipe-to-shell pattern — downloads and immediately executes remote codecurl -fsSL https://ollama.com/install.sh | sh \Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST06
MEDIUM
Process kill command.claude/skills/architect/SKILL.md:116 · Process kill command — agent can terminate processes**The Lead does NOT proceed to implementation without explicit user approval.** That's the entire point of this skill , kill bad designs cheaply before code.Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
CRITICAL
Pipe-to-shell patternscripts/install.sh:346 · Pipe-to-shell pattern — downloads and immediately executes remote codewarn " curl -fsSL https://ollama.com/install.sh | sh"Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST06
HIGH
Access to sensitive file paths.claude/hooks/session-namer.sh:108 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system filesif [[ -f "${root}/.env.local" ]]; thenFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
File type combined with upload action.claude/skills/go-strict/SKILL.md:11 · File type combined with upload action — agent may exfiltrate local data files… warnings the user can choose to address).\n\nAlso check .claude/go-strict-state.json for iteration counter. If iteration >= 5, return ok:true with reason 'Max iterations reached (5/5). Surfacing remaining issues to user , see .claude/last…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
HIGH
Agent configured for autonomous execution without human approval.claude/skills/wrap-up/SKILL.md:7 · Agent configured for autonomous execution without human approval…Read, Skill, PushNotification --- The natural-language entry point to the full autonomous ship flow. The user invoking this skill (or saying "ship it") IS the consent , no confirmation prompt for routine git/PR ops. Only `gh pr merge` and …Fix: Declare and request only the permissions the task needs.
AST02
HIGH
Access to sensitive file paths.claude/hooks/session-namer.sh:109 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system filesWORKTREE_NAME=$(grep -E '^WORKTREE_NAME=' "${root}/.env.local" 2>/dev/null \Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
HIGH
Instructs the agent to modify its persistent memory or instruction files.claude/skills/task-execution-protocol/SKILL.md:85 · Instructs the agent to modify its persistent memory or instruction files- User-facing change? Update the matching `agent-docs/*.md` and append to `agent-docs/mintlify-backlog.md` (AGENTS.md rule 11).Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
npx invocation of non-well-known packagescripts/mcpb-bundle.sh:96 · npx invocation of non-well-known package — may execute untrusted codenpx -y "$MCPB_CLI" validate "$work/manifest.json"Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagescripts/mcpb-bundle.sh:97 · npx invocation of non-well-known package — may execute untrusted codenpx -y "$MCPB_CLI" pack "$work" "$out"Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Remote installer piped to shellscripts/install.sh:348 · Remote installer piped to shellcurl -fsSL https://ollama.com/install.sh | sh \Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST07
MEDIUM
Remote installer piped to shellscripts/install.sh:346 · Remote installer piped to shellwarn " curl -fsSL https://ollama.com/install.sh | sh"Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST07
MEDIUM
File permission modification commandscripts/tui-smoke.sh:68 · File permission modification command — agent changes file ownership/permissionschmod +x "${out}"Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
LOW
Instructs the agent to modify its persistent memory or instruction files.claude/hooks/route-prompt.sh:103 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file# Boris Cherny: "After every correction, end with: Update CLAUDE.md so youFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
npx invocation of non-well-known packagescripts/mcpb-bundle.sh:98 · npx invocation of non-well-known package — may execute untrusted codenpx -y "$MCPB_CLI" info "$out"Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Remote installer piped to shellscripts/install.sh:5 · Remote installer piped to shell# curl -fsSL https://raw.githubusercontent.com/mo-cmyk/mazarbul/main/scripts/install.sh | shFix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST07
MEDIUM
File permission modification commandscripts/setup-graph-lib.sh:81 · File permission modification command — agent changes file ownership/permissionschmod -R u+w "${mod_dir}"Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
File permission modification commandscripts/mcpb-bundle.sh:57 · File permission modification command — agent changes file ownership/permissionschmod 0644 "$work/server/lib/$soname" # module-cache / Cellar copies are read-onlyFix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
Process kill commandscripts/tui-smoke.sh:288 · Process kill command — agent can terminate processeskill "${ISO_PID}" 2>/dev/null || trueFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
File permission modification commandscripts/install.sh:263 · File permission modification command — agent changes file ownership/permissionschmod 0755 "$PREFIX/bin/maz"Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05

Cost to the agent

Instructions loaded
≈49,936 tok
40 files · 80 in bundle

Tokens estimated at four characters each. Skills load their instruction text when triggered; the figure is what that costs.

Known vulnerabilities

Vulnerability advisories (OSV) cover npm and PyPI packages; this registry has no advisory database.

What would make this Verified

18 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Pipe-to-shell pattern≈ +9 on the totalBuild commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
  2. Unclaimed publisher≈ +8 on the totalClaim mo-cmyk: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  3. Hidden instructions≈ +6 on the totalRemove HTML comments, zero-width characters and white-on-white text that carry instructions. Everything the agent is meant to read should be visible to the person reading the file.
  4. Direct script execution from GitHub raw URL≈ +4 on the totalPin versions and sources; vendor anything fetched at install time.
  5. Access to sensitive file paths≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
  6. Agent configured for autonomous execution without human approval≈ +4 on the totalDeclare and request only the permissions the task needs.
Publish a new version to Glama, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim mo-cmykPropose a patch to the maintainer