Agent Economy Report

Skill code audit

CAUTION version 1.0.0 · audited 2026-09-17

auto-updater @maximeprades

This is a Markdown instruction package that configures a daily auto-update routine for Clawdbot and installed skills. It schedules a cron job and runs commands such as npm/pnpm/bun update -g clawdbot@latest, clawdbot update, clawdbot doctor --yes, and clawdhub update --all, then reports a formatted summary. It can write a helper script and log file under ~/.clawdbot and suggests using sudo if an update hits a permission error. The package contains no obfuscated code, hidden network endpoints, credential exfiltration, or payment/wallet redirection; the referenced domains are public documentation and the ClawHub registry. The main concern is that it replaces Clawdbot and every installed skill with the latest versions on a schedule without per-update approval.

Declared purposeAutomatically update Clawdbot and all installed skills once daily via cron, apply updates, and message the user with a summary of what changed.
Observed behaviorThe code instructs the agent to detect the Clawdbot install type, create an optional update script, add a daily cron job using clawdbot cron add, run package and skill update commands, run clawdbot doctor --yes, write logs, and format a summary. It also provides troubleshooting guidance, including a suggestion to use sudo for EACCES errors.
Verdict, rules onlyCAUTION · uses sudo, rm -rf or dd
Verdict, AICAUTION
Final verdictCAUTION (never better than either pass)

Capabilities

capabilityvs purposedetail
networkneededFetches updates from npm/pnpm/bun and ClawdHub via clawdhub update --all; the only domains present in the files are docs.clawd.bot and clawhub.ai.
shellneededRuns shell commands such as npm, pnpm, bun, clawdbot, clawdhub, mkdir, tee, and date as part of the update procedure.
filesystem_readneededChecks install type and versions with commands like npm list -g clawdbot and directory tests for ~/.clawdbot/.git or /opt/clawdbot/.git/config.
filesystem_writeneededWrites an optional helper script at ~/.clawdbot/scripts/auto-update.sh and logs to ~/.clawdbot/logs/auto-update.log; update commands also modify installed packages and skills.
install_packagesneededUpdates the global Clawdbot package with npm/pnpm/bun update -g clawdbot@latest and updates all skills via clawdhub update --all.
persistenceneededAdds a daily Clawdbot cron job named Daily Auto-Update so the update routine runs on a schedule.

Findings

severityfinding
infoDeclared purpose
This is the package's stated purpose and matches the observed update and cron instructions.
SKILL.md:3 Automatically update Clawdbot and all installed skills once daily. Runs via cron, checks for updates, applies them, and messages the user with a summary of what
lowAdds a daily cron job
The skill schedules a recurring daily job using clawdbot cron add. This is persistent scheduling and is central to the declared purpose.
SKILL.md:34 --cron "0 4 * * *"
mediumInstalls latest Clawdbot automatically
Runs npm update -g clawdbot@latest, with pnpm and bun variants nearby. This can replace the installed Clawdbot package with any newly published version without per-update approval.
references/agent-guide.md:44 npm update -g clawdbot@latest 2>&1 | tee -a "$LOG_FILE"
mediumUpdates all installed skills automatically
This command can replace every installed skill with new registry versions, so third-party skill code may change without fresh review.
SKILL.md:71 clawdhub update --all
lowRuns doctor with automatic confirmation
The update procedure runs clawdbot doctor --yes automatically, which can apply migrations and system changes without prompting.
references/agent-guide.md:56 clawdbot doctor --yes 2>&1 | tee -a "$LOG_FILE" || true
infoOptional helper script creation
The guide tells the agent it may create a local shell script for the update routine.
references/agent-guide.md:23 For complex setups, create a helper script at `~/.clawdbot/scripts/auto-update.sh`:
infoWrites a log file
The procedure writes update logs to ~/.clawdbot/logs/auto-update.log using tee and redirects.
references/agent-guide.md:29 LOG_FILE="${HOME}/.clawdbot/logs/auto-update.log"
lowSuggests sudo for permission errors
The troubleshooting guidance tells the agent to suggest sudo when permission errors occur. It does not directly run sudo.
references/agent-guide.md:150 - `EACCES`: Permission denied → suggest `sudo` or fixing permissions
lowPublisher discloses the main risk
The skill card openly states that daily auto-updates can replace Clawdbot and every skill without fresh approval.
skill-card.md:24 Risk: Daily auto-updates can replace Clawdbot and every installed skill without fresh approval.
infoDocumented network reference
Network references are visible documentation and registry links. No hidden IP addresses or encoded domains were found.
SKILL.md:147 - [Clawdbot Updating Guide](https://docs.clawd.bot/install/updating)

0 AI finding(s) were dropped because their file, line or quote did not match the package.

Network destinations in the code

domainservice ratingwhere
clawhub.ainot in the service indexskill-card.md:9
docs.clawd.botnot in the service indexSKILL.md:147

Other facts

Binaries invokedbash, git, npm, sudo
Environment variables readCLAWDBOT_VERSION_AFTER, CLAWDBOT_VERSION_BEFORE, HOME, LOG_FILE, SKILL_OUTPUT
Hard-coded walletsnone
Pipes a download to a shell0
eval / exec / subprocess0
base64 blobs0
File writes15
Persistence0

Files audited

filelinessha256
SKILL.md1494142916e146caa90…
_meta.json6d9d38e5d386deb7a…
references/agent-guide.md152006191441b38b44f…
references/summary-examples.md1099f0c386e2fee1fe5…
skill-card.md567fb91f1f9c9d20e8…

For agents

JSON: https://agenteconomy.report/k/auto-updater.audit.json · badge: https://agenteconomy.report/k/auto-updater.audited.svg · skill rating: /k/auto-updater · commission an audit of any skill: US$ 29 per version.

How this is computed

The complete published package of this exact version was downloaded from the registry and read statically; nothing was executed. A deterministic pass extracts network destinations, binaries, environment variables, writes, install commands, obfuscation markers, subprocesses and wallets, each with file and line. An AI then reads the whole package with those facts and writes the summary, the capabilities and the findings under a strict schema; every finding it produces must cite a file, a line and the exact text, or it is dropped. The final verdict is the worst of the two passes. The audit does not change the skill's trust tier (policy); the author may respond through the dispute channel and the response is published here. Commissioned by: the Agent Economy Report (free program: the 150 most downloaded skills, September 2026).