BC jobwatch
Autonomous job-market watcher for OpenClaw. On a cron schedule it monitors the career pages of companies the user has EXPLICITLY configured, judges each posting against the user's own job profile (visa / seniority / red lines) with an LLM, pushes strong matches, digests the rest daily, and archives postings into a knowledge base; it also tracks the user's application status and answers questions about their own watched jobs. PRIVACY & CAPABILITIES (see the Privacy & Data Flow section): this skill collects and stores a personal job-seeking profile (resume text, visa needs, seniority, red lines); sends watched URLs and job-description text to third-party services (Firecrawl / Jina scrapers, an OpenAI-compatible LLM endpoint, and a 2brain knowledge base); reads host OpenClaw / Telegram credentials only when the user opts in (JOBWATCH_ALLOW_HOST_CREDS=1); and registers a recurring cron job — every autonomous action happens only after an explicit onboarding consent step. Use ONLY when the user clearly wants automated job monitoring for their own search — not on casual mention of jobs. Trigger when the user asks to set up or run job monitoring, to watch specific companies' careers pages, reports their own application update, or asks about their own watched/matched jobs. Triggers — imperative requests only; a bare mention of jobs, careers or 求职 must NOT activate this skill: 'set up jobwatch for me', 'start job monitoring for my search', 'watch <company>'s careers page for me', 'I applied to <specific job>', 'jobwatch top jobs this week', '帮我设置求职监控', '开启求职监控', '帮我盯 <公司> 的岗位', '我投了 <职位>'. If the user is only discussing or asking about jobs in general, do not activate. Before the first onboarding question — and before any profile data is collected, any credential is read, or any cron job is registered — state what the skill will collect and send, and get an explicit yes.
Autonomous job-market watcher for OpenClaw.
As a process C 62/100 · Has gaps — weak spots: result and completion, running it twice
What is at stake
Medium-severity findings: the skill is probably honest, but read what alarmed the scanner.
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
- Shorten the description to 1024 characters.
- 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 · 2
✓ No critical or high findings
Medium and low: 2
-
medium Exfiltration
net-redirectable-api-keyscripts/common.py:342Helper sends the API key to a host configured by an environment variable — the key can be redirected to another serverAPI key + configurable base URL from environment
-
low Exfiltration
exfil-webhook-urlscripts/notify_telegram.py:32Webhook / callback URL commonly used for exfiltration (verify the destination) (the skill's own vendor host; quoted — discussed, not commanded)f"https://api.telegram.org/bot{token}/sendMessage",vendor-hostquoted
Files scanned: 25. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- error
description-longdescription is 1893 chars, limit 1024 - note
description-budgetdescription takes 1893 of the ~15000-char shared budget for all skills
Process rating: all ten parameters 62/100
- 0Result and completion. Does not say what the result is
- 30Running it twice. 26 mutating operations with no state check
- 50Failures and branches. 0 branches, has a failure section
- 60Tools and files. Uses tools (bash, web, python) that frontmatter does not declare
- 70When it triggers. States when to use, but not when not to
- 70Inputs and preconditions. Inputs and preconditions are listed
- 70Execution cost. Instruction body is 4961 tokens
- 100Steps. 41 steps
- 100Consistency. Name and required fields are in place
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- high The skill tells the model to perform an irreversible action with no human approval
- low The response is described with custom markup (12 tags): a typed call is more reliable
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 1893: 120–800 characters recommended
- +3Output format is not stated: the model decides each time
- -35 of 14 scripts are never mentioned in SKILL.md
- +1No license
- +2Single-language instructions
- +4Structure: 11 headings
- +3Step-by-step instructions: 41 items
- +4Has examples (2 code blocks)
- +4Reference files are cited in the instructions (1 of 1)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 54.