CC codex-autopilot
Multi-model AI coding automation system with intelligent task routing and built-in CI/CD. Watchdog-driven loop that orchestrates Codex (backend) and Gemini (frontend) sessions in tmux, auto-routes tasks by type, manages context compaction, runs incremental code reviews, dispatches tasks from a priority queue, and includes a test-agent that auto-detects failures, enqueues fixes, and ratchets coverage. Use when managing multiple concurrent AI coding sessions, automating development workflows, orchestrating parallel AI-assisted coding across projects, routing frontend vs backend tasks to different models, or running continuous testing with auto-fix. Triggers: autopilot, watchdog, codex automation, tmux codex, multi-project codex, auto-nudge codex, codex session management, gemini frontend, multi-model routing, ai task routing, ci cd, test agent, coverage ratchet.
As a process C 55/100 · Has gaps — weak spots: result and completion, when it triggers, inputs and preconditions
What is at stake
The skill contains fragments that, in the wrong hands, cost money or data. Below: what the installer risks and what the author should do.
The skill contains commands that delete files, rewrite disks or execute code fetched from the network. The agent may run them without asking if it believes the instructions require it.
Replace destructive commands with safe equivalents that ask for confirmation, scope them to one folder, and stop piping curl into a shell: pin a version and a checksum.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The instructions or scripts send data out: environment variables, keys, file contents, chat history. You may never notice, because the agent performs the upload as "part of the task".
If the upload is not needed for the task, remove it: catalogs flag such skills and delist them. If it is needed, name the destination explicitly, say what leaves the machine, and give the user a switch.
How to improve
- Address the high-severity findings: each costs 18 safety points. If one is a false positive, add the rule id to guard.allow in spec.yaml.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 9
-
high Dangerous commands
cmd-persistenceinstall.sh:243Persistence mechanism (cron / launchd / scheduled task / autorun registry)launchctl load "$PLIST_PATH"
Medium and low: 8
-
medium Dangerous commands
cmd-persistenceinstall.sh:24Persistence mechanism (cron / launchd / scheduled task / autorun registry) (string literal in code, not executed)PLIST_PATH="${HOME}/Library/LaunchAgents/${PLIST_NAME}.plist"code literal -
medium Dangerous commands
cmd-persistenceinstall.sh:264Persistence mechanism (cron / launchd / scheduled task / autorun registry) (string literal in code, not executed)local p="${HOME}/Library/LaunchAgents/${name}.plist"code literal -
medium Exfiltration
exfil-webhook-urllib/telegram_bot.py:69Webhook / callback URL commonly used for exfiltration (verify the destination) (quoted — discussed, not commanded)self.api_base = f"https://api.telegram.org/bot{bot_token}"quoted -
medium Exfiltration
exfil-webhook-urllib/telegram_notifier.py:63Webhook / callback URL commonly used for exfiltration (verify the destination) (quoted — discussed, not commanded)self.api_base = f"https://api.telegram.org/bot{bot_token}"quoted -
medium Exfiltration
net-credential-usescripts/autopilot-lib.sh:709Credential used in a network call (verify the destination is the intended service)curl -s -X POST "https://api.telegram.org/bot${LIB_TG_TOKEN}/sendMessage" \ -
low Dangerous commands
cmd-background-processREADME_zh.md:336Starts a background / autostarted processnohup bash scripts/watchdog.sh &
-
low Dangerous commands
cmd-background-processREADME.md:232Starts a background / autostarted processnohup bash scripts/watchdog.sh &
-
low Exfiltration
exfil-webhook-urlscripts/autopilot-lib.sh:709Webhook / callback URL commonly used for exfiltration (verify the destination) (placeholder value)curl -s -X POST "https://api.telegram.org/bot${LIB_TG_TOKEN}/sendMessage" \placeholder
Files scanned: 58. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
✓ No remarks against the Agent Skills spec
Process rating: all ten parameters 55/100
- 0Result and completion. Does not say what the result is
- 0Inputs and preconditions. Does not say what the process needs to start
- 20When it triggers. No condition that starts the skill
- 30Running it twice. 13 mutating operations with no state check
- 60Tools and files. Uses tools (bash, python) that frontmatter does not declare
- 100Steps. 33 steps
- 100Failures and branches. 1 branches, has a failure section
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 2197 tokens
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +4Description does not say when NOT to use the skill (false activations)
- +3Description length 872: 120–800 characters recommended
- +3Output format is not stated: the model decides each time
- -44 reference files, but SKILL.md never points to them: the model will not open them
- -34 of 26 scripts are never mentioned in SKILL.md
- +1No license
- +2Single-language instructions
- +4Structure: 29 headings
- +3Step-by-step instructions: 33 items
- +4Has examples (16 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 74.