CareerSignal OS
CareerSignal is a portable professional-intelligence skill bundle for Claude, Claude Code, Codex, ChatGPT-compatible plugin environments, and agents that read SKILL.md.
Core loop:
Research → Learn → Build → Prove → Teach → Measure → Improve
Zero-setup mode
CareerSignal works without paid connectors.
/CSstart initializes the user's professional context once./CSpost researches and builds the best post for that day.- The result is returned for manual copy/paste.
- If image generation exists, use it when useful.
- If it does not, return the exact visual/carousel brief.
- Optional connectors may remove manual steps, but they never unlock the core intelligence.
Command routing
Read the relevant focused skill:
/CSstart → skills/onboarding/SKILL.md/CSpost, /CSideas, /CSseries, /CSvisual, /CSpoll, /CSreply, /CSrepost → skills/daily-post/SKILL.md/CSbrief → skills/research-brief/SKILL.md/CSprofile → skills/profile-audit/SKILL.md/CScareer, /CSgaps → skills/career-gap/SKILL.md/CSproject → skills/build-project/SKILL.md/CSlearn, /CStest → skills/learn-topic/SKILL.md/CSaudit → skills/post-audit/SKILL.md/CShistory → skills/performance-learning/SKILL.md/CSsettings, /CSexport, /CSreset → skills/user-context/SKILL.md/CSconnectors → skills/connector-setup/SKILL.md
Natural-language equivalents also work.
First-run rule
If a personalized workflow is requested before setup, run onboarding first and then continue the original request.
Ask once for:
- CV/resume or LinkedIn profile;
- desired roles/topics;
- optional GitHub/portfolio;
- optional old posts/screenshots;
- preferred language;
- manual or connected workflow.
Do not ask a long questionnaire.
Trust model
Every public claim is constrained by:
- evidence confidence: GREEN / YELLOW / RED;
- knowledge boundary: BUILT / USED_PROFESSIONALLY / USED_IN_PROJECT / STUDIED / LEARNING / READ_ABOUT / UNKNOWN;
- privacy: public / ask_first / private.
Never turn learning into expertise. Never invent metrics, clients, seniority, production use, or results.
Daily-post rule
Internally:
- research current developments and real practitioner problems;
- generate multiple candidate problems;
- score for freshness, usefulness, evidence, fit, novelty, memorability, and connection value;
- choose domain, category, audience, and format independently;
- check counterarguments and whether the user can defend the post;
- use BUILD instead when a small experiment would create stronger original evidence;
- humanize, fact-check, privacy-check, and originality-check;
- create the best asset when useful.
User-visible /CSpost output should normally be only the final post plus its selected asset.
Connector principle
Connectors remove work. They do not unlock intelligence.
Every connector-assisted workflow must have a manual fallback.
External write actions require explicit user approval.
Never silently publish.
Extended operational commands
/CSinterview — build Story Bank / gather missing personal evidence/CScomment — draft a substantive comment/CSplan — build a flexible content plan/CSrepurpose — rebuild existing content for a new professional use/CShook — analyze/improve the opening/CSengage — analyze useful audience engagement/CSmonitor — review comments/replies/outcomes for follow-ups/CSteam — team/employee-advocacy mode
Reliability commands
/CShelp — concise command help/CSdoctor — health/capability check/CSprivacy — privacy/state inspection/CSrestore — restore a validated backup
External write firewall
Any external write must be gated by lib/approval_firewall.py when executable adapters are used. Explicit user approval must bind to the exact payload being written.
1---2name: career-signal-os3description: Research-driven professional intelligence and content OS. Personalize once from a CV/profile and optional projects, then use /CSpost to research current problems in the user's field, choose the strongest evidence-backed topic and format, create a humanized LinkedIn-ready post or learning asset, and keep free/manual fallbacks when connectors are unavailable. Also supports /CSstart, /CSbrief, /CSprofile, /CScareer, /CSproject, /CSlearn, /CStest, /CSaudit, /CShistory, /CSsettings, /CSconnectors, /CSideas, /CSseries, /CSvisual, /CSpoll, /CSreply, and /CSrepost.4---56# CareerSignal OS78CareerSignal is a portable professional-intelligence skill bundle for Claude, Claude Code, Codex, ChatGPT-compatible plugin environments, and agents that read `SKILL.md`.910Core loop:1112**Research → Learn → Build → Prove → Teach → Measure → Improve**1314## Zero-setup mode1516CareerSignal works without paid connectors.1718- `/CSstart` initializes the user's professional context once.19- `/CSpost` researches and builds the best post for that day.20- The result is returned for manual copy/paste.21- If image generation exists, use it when useful.22- If it does not, return the exact visual/carousel brief.23- Optional connectors may remove manual steps, but they never unlock the core intelligence.2425## Command routing2627Read the relevant focused skill:2829- `/CSstart` → `skills/onboarding/SKILL.md`30- `/CSpost`, `/CSideas`, `/CSseries`, `/CSvisual`, `/CSpoll`, `/CSreply`, `/CSrepost` → `skills/daily-post/SKILL.md`31- `/CSbrief` → `skills/research-brief/SKILL.md`32- `/CSprofile` → `skills/profile-audit/SKILL.md`33- `/CScareer`, `/CSgaps` → `skills/career-gap/SKILL.md`34- `/CSproject` → `skills/build-project/SKILL.md`35- `/CSlearn`, `/CStest` → `skills/learn-topic/SKILL.md`36- `/CSaudit` → `skills/post-audit/SKILL.md`37- `/CShistory` → `skills/performance-learning/SKILL.md`38- `/CSsettings`, `/CSexport`, `/CSreset` → `skills/user-context/SKILL.md`39- `/CSconnectors` → `skills/connector-setup/SKILL.md`4041Natural-language equivalents also work.4243## First-run rule4445If a personalized workflow is requested before setup, run onboarding first and then continue the original request.4647Ask once for:481. CV/resume or LinkedIn profile;492. desired roles/topics;503. optional GitHub/portfolio;514. optional old posts/screenshots;525. preferred language;536. manual or connected workflow.5455Do not ask a long questionnaire.5657## Trust model5859Every public claim is constrained by:60- evidence confidence: GREEN / YELLOW / RED;61- knowledge boundary: BUILT / USED_PROFESSIONALLY / USED_IN_PROJECT / STUDIED / LEARNING / READ_ABOUT / UNKNOWN;62- privacy: public / ask_first / private.6364Never turn learning into expertise. Never invent metrics, clients, seniority, production use, or results.6566## Daily-post rule6768Internally:691. research current developments and real practitioner problems;702. generate multiple candidate problems;713. score for freshness, usefulness, evidence, fit, novelty, memorability, and connection value;724. choose domain, category, audience, and format independently;735. check counterarguments and whether the user can defend the post;746. use BUILD instead when a small experiment would create stronger original evidence;757. humanize, fact-check, privacy-check, and originality-check;768. create the best asset when useful.7778User-visible `/CSpost` output should normally be only the final post plus its selected asset.7980## Connector principle8182**Connectors remove work. They do not unlock intelligence.**8384Every connector-assisted workflow must have a manual fallback.85External write actions require explicit user approval.86Never silently publish.8788## Extended operational commands8990- `/CSinterview` — build Story Bank / gather missing personal evidence91- `/CScomment` — draft a substantive comment92- `/CSplan` — build a flexible content plan93- `/CSrepurpose` — rebuild existing content for a new professional use94- `/CShook` — analyze/improve the opening95- `/CSengage` — analyze useful audience engagement96- `/CSmonitor` — review comments/replies/outcomes for follow-ups97- `/CSteam` — team/employee-advocacy mode9899## Reliability commands100- `/CShelp` — concise command help101- `/CSdoctor` — health/capability check102- `/CSprivacy` — privacy/state inspection103- `/CSrestore` — restore a validated backup104105## External write firewall106107Any external write must be gated by `lib/approval_firewall.py` when executable adapters are used. Explicit user approval must bind to the exact payload being written.
venilkukadiya52