Borrowing it
Nothing to install: this file belongs to colophon-group/jobseek. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/colophon-group/jobseek/main/.agents/skills/jobseek-label-daily/SKILL.mdgit clone --depth 1 https://github.com/colophon-group/jobseekWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/colophon-group/jobseek/jobseek-label-daily)<a href="https://agentmods.dev/skills/colophon-group/jobseek/jobseek-label-daily"><img src="https://agentmods.dev/badge/skills/colophon-group/jobseek/jobseek-label-daily/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/colophon-group/jobseek/jobseek-label-daily"><img src="https://agentmods.dev/badge/skills/colophon-group/jobseek/jobseek-label-daily.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00075 | $0.01740 |
| Opus 5 | $0.00037 | $0.00870 |
| Sonnet 5 | $0.00015 | $0.00348 |
| Haiku 4.5 | $0.00007 | $0.00174 |
Grade A, and why
jobseek-label-daily scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 13d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 185 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Jobseek Label Daily
Overview
This routine samples public job postings from the last 24 hours, labels them with Codex project custom agents, validates the outputs, merges accepted records, and uploads accepted gold data to viktoroo/jobseek-postings-labelled.
The Python labeller code is deterministic orchestration only: database reads, Jinja task rendering, JSON Schema/custom validation, merge, QA, and HuggingFace upload. Do not add provider SDK calls or call OpenAI or Anthropic endpoints from apps/crawler/src/labeller. LLM judgment happens through the already-running, subscription-backed Codex session and project custom agents configured in the active harness; the repo mirrors durable subagent contracts under .agents/labeller/.
Agents
Use these project custom agents:
jobseek-labeller-normalizer: GPT-5.6 Luna at low reasoning; raw HTML task input ->normalized.html.jobseek-labeller-splitter: GPT-5.6 Luna at medium reasoning;split_sectionstask input ->split-out.json.jobseek-labeller-extractor: GPT-5.6 Terra at high reasoning;extract_alltask input ->extract-all-out.json.
The committed .codex/agents/*.toml files pin these settings. Do not let the
labeller agents inherit the Sol/high orchestrator default: normalization and
splitting are clear, repeatable tasks, while combined extraction needs more
semantic judgment without requiring the flagship model.
Each agent invocation message must be exactly:
INPUT: <rendered-input-path>
OUTPUT: <output-path>
The Jinja task prompts in apps/crawler/src/labeller/prompts/tasks/ own task-specific rules and schema phrasing. Do not duplicate or rewrite those prompts in the orchestrator.
Workflow
Run commands from apps/crawler with uv run labeller .... Default RUN_DATE is today; default accepted-record target is 10 unless the user provides overrides. Start with SAMPLE_SIZE=10; if QA rejections leave fewer than 10 accepted local records for the date, sample additional candidates and continue until the accepted target is met or the sampling pool is exhausted.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 13d ago First seen · 185 lines · 75 tokens per session scan A b97b2117252a
jobseek-label-daily is a skill published in the GitHub repository colophon-group/jobseek (191 stars, last pushed yesterday), licensed MIT. It adds 75 tokens to every session and 1,740 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
job-coach
The persona-driven orchestrator for the whole job search. Treats the user like a real career-coaching client — opens with intake (goals, companies, timeline, geography, constraints), researches the market visibly, presents a ranked slate of next moves with suggestions, and chains into the verb skills (/job-apply…
job-interview
Prep the candidate for a specific upcoming interview — research the company + role + likely questions, generate a STAR answer bank tied to the candidate's actual projects (not generic), surface smart questions to ask, and flag red flags. Proactively invoke this skill (do NOT answer conversationally) when the user says…
job-cv
Tailor the candidate's base CV for one specific job posting and render an ATS-friendly PDF. Reads cv.md (or extracts a CV from config/candidate-profile.md), pulls the job description from a URL or pasted text, has Claude rewrite — never invent — bullets to match the JD's keywords/priorities, gets explicit user…
job-evaluate
Auto-pipeline for a single job posting. Pastes a URL → fetches the JD → scores it across the 7-block A-G rubric → writes a structured evaluation report → generates a tailored ATS-friendly PDF CV → appends a row to application-tracker.csv. Career-ops's killer demo, ported to our skill pack. Proactively invoke this…
job-followup
Check outreach-log.csv for cold emails that haven't replied, compute urgency (7-day cadence, max 2 follow-ups per contact), draft and send follow-ups one at a time. Proactively invoke this skill (do NOT answer conversationally) when the user asks "who should I follow up with", "any contacts gone cold", "run the day-7…
job-outreach
Research a company or hiring manager, draft a personalized cold email in chat, get approval, send via local msmtp, log for day-7 follow-up tracking. Proactively invoke this skill (do NOT answer conversationally) when the user mentions reaching out to a hiring manager / VP / recruiter / director, pastes a LinkedIn…