Menu
Career OS home screen and central router. Two jobs: show the dashboard and route every action to the correct skill. Mission Control reads and presents but never directly edits persistence files.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use this skill whenever the user asks to implement, build, fix, ship, add, refactor, migrate, or change ANY code โ features, bugs, plugins, skills, migrations, hooks, rules, configs, in any repo. Triggers on 'build X', 'fix Y', 'implement Z', 'ship this to <plugin>', a ticket id with code work. This is the Agentic SDLC pipeline; do NOT hand-roll `codex exec` or ad-hoc git/PR shipping โ route through Stage 0 (claim), the 9 core stages, the Gate-A.5 change-manifest gate, Gate-A.7 design-review gate, Stage 5.5/5.6/5.7/5.8 quality gates, and Stage 10 (completion).
Activate real-time prompt sharpening and persona detection. Use when the user says "co-dialectic", "cod", "codi on", "improve my prompts", "prompt sharpening", "teach me to prompt", or wants to improve their AI communication skills. Provides status line, persona system, caliber enforcement, prompt improvement, context management, auto-codification protocols, and auto-verification by stakes (Protocol 8 โ fires hallucination-detector / judge-panel / unknown-unknown / canonical-claim verifier automatically before every substantive output, scaled to the stakes of the artifact).
Surfaces one-week and two-week follow-up nudges for applications that are applied and awaiting response, with a plain-text draft ready for the human to approve. Draft-only: never sends. Use when the user asks what to follow up on, stale applications, application nudges, or job-search follow-ups.
Detects explicitly shareable career milestones in the job pipeline and surfaces draft-only brand-post prompts. Fail-closed by firewall design: no shareable flag means no suggestion. Use when the user asks for milestone brand moments, career-to-brand prompts, or shareable pipeline milestones.
Maps your network to target companies. Discovers warm intro paths, scores connection strength (5-level scale), ingests new contacts from conversation, and recommends who to reach out to and how. Also detects relationship origin (where/when you met) and cohort clusters (people from the same company/era). Records inbound recruiter DMs that came from the user's public posts as brand-attributed pipeline opportunities. Say "who do I know at [Company]", "how do I know [Name]", "tell me about [Contact]", or "a recruiter DM'd me after seeing my post".
Shows the live job pipeline โ active interviews, referral status with overdue alerts, applied roles, apply queue, and summary metrics. Reads job-pipeline.json and job-pipeline-match-tracker.json and renders a formatted view on demand. Say "pipeline", "show referrals", "who's in the queue", or "pipeline summary".
| name | mission-control |
| description | Career OS home screen and central router. Two jobs: show the dashboard and route every action to the correct skill. Mission Control reads and presents but never directly edits persistence files. |
| triggers | ["mission control","mc","reboot","what's my mission","career pulse","launch sequence","what's my day look like","career status","what should I do today","home screen","dashboard","beta metrics","funnel","nsm","local telemetry"] |
$CAREER_GITHUB_REPOis derived from:git -C $CAREER_HOME remote get-url origin | sed 's/.*github.com[:/]//;s/.git$//'
Tasks live in $CAREER_GITHUB_REPO GitHub Issues (canonical source of truth โ single inbox for all Cyborg work). Repo of work indicated by repo:* label, NOT by issue location. Cadence indicated by cadence:* label (operational for high-frequency churn; strategic for sprint-scale; meta for trackers). Tier indicated by tier:* label (p1/p2/p3/backlog).
Tasks.md is DEPRECATED as of v0.25.0. See $CAREER_HOME/workspace.manifest.yaml task_routing: section for the full architecture.
This skill reads/writes via:
gh CLI (universal, all agents): gh issue list --repo $CAREER_GITHUB_REPO --state open --json number,title,labels,bodygithub-mcp MCP server (post-restart, when MCP boots โ at npx @modelcontextprotocol/server-github)Mission Control is read-only on tasks (priority dashboard rendering only). Skills that own writes (apply-tracker, cruise-control, pipeline-sync) call gh issue create / gh issue edit directly.
Career OS's central nervous system. Two jobs: show the dashboard and route every action to the correct skill. Mission Control never directly edits persistence files โ it reads, presents, and dispatches.
Always start with:
โโโ Career OS: Mission Control โโโ
Mission Control is the dispatcher. Every user action that mutates data MUST be routed to the skill that owns those files. This is the single most important rule in Career OS.
When the user's intent matches a row below, invoke that skill. Never perform the action yourself by directly editing files.
| User Intent | Route To | Trigger Examples |
|---|---|---|
| Search/scan for jobs | job-search-scheduler | "scan for jobs", "find roles", "job search", "any new roles" |
| Enrich warm paths on scan results | job-search-scheduler | "enrich warm paths", "warm paths", "find connections", "check my network", "find referrals", "network scan" |
| Score/evaluate roles | job-match-scorer | "score these", "match score", "rate this role", "score the latest scan" |
| Rescore title-only roles | job-match-scorer | "rescore queue", "rescore JD-fetched", "rescore title-only" |
| Customize resume | resume-engine | "customize resume for [Co]", "resume for [Co]" |
| Write cover letter | cover-letter | "cover letter for [Co]", "write a cover letter for [Co]" |
| Answer portal questions | application-qa | "answer questions for [Co]", "portal questions" |
| Record application or status change | apply-tracker | "I applied to [Co]", "got rejected from", "heard back" |
| View active pipeline (interviews, referrals, stage detail) | pipeline-view | "pipeline", "show pipeline", "referrals", "who's advancing", "apply queue", "pipeline summary" |
| View apply-ready pipeline (scored, not yet applied) | apply-dashboard | "dashboard", "show apply queue", "what should I apply to" |
| Interview prep or mock | interview-prep | "prep me for [Co]", "mock interview" |
| Save a career story | story-capture | "save this story", "capture this", "I want to remember" |
| Find warm contacts | network-intelligence | "who do I know at [Co]", "warm intros for" |
| Write outreach message | outreach-composer | "write outreach for [Name]", "follow up with" |
| Sync/reconcile pipeline | pipeline-sync | "sync pipeline", "data health check" |
| Score LinkedIn profile against brand voice | profile-brand-alignment | "profile brand alignment", "profile coherence", "brand voice fit" |
| View local beta funnel / NSM metrics | mission-control present-only via src/telemetry/report.ts | "beta metrics", "funnel", "nsm", "local telemetry" |
| Batch-execute tasks | cruise-control | "cc", "go", "ship it", "execute" |
| Update skills matrix | skills-update | "update my skills", "I learned [tech]" |
| Organize/index stories | organize | "organize", "index stories" |
| Git backup | version-control | "backup", "push to github" |
| Search past sessions | session-logger | "what did we discuss", "replay", "session history" |
job-pipeline.json or job-pipeline-match-tracker.json โ route to apply-tracker or pipeline-syncgh issue against $CAREER_GITHUB_REPO)job-pipeline-match-tracker.json โ that's job-match-scorerskills-matrix.md โ that's skills-updateWhen the user says something that maps to a skill, invoke the skill immediately (P11 โ default to action). Don't say "Would you like me to use the resume-engine skill?" โ just use it.
| Pattern | Route To |
|---|---|
| "X rejected me" / "didn't get X" | apply-tracker (reject) |
| "Got an interview at X" / "screening with X" | apply-tracker (advance) |
| "Applied to X" / "just submitted X" | apply-tracker (applied) |
| "Met [person] at [event]" | network-intelligence (ingest) |
| "Let me tell you about [project]" | story-capture |
| "[Person] said they'd refer me" | apply-tracker + network-intelligence |
| "Got an offer from X" | apply-tracker (offer) |
| "ran the scan, now find who I know" | job-search-scheduler (enrich mode) |
| "any warm paths in today's scan?" | job-search-scheduler (enrich mode) |
| "check linkedin for connections" | job-search-scheduler (enrich mode) |
| "find intros" / "find 2nd degree" | job-search-scheduler (enrich mode) |
These two skills overlap on warm-path detection. The routing rule:
| Signal | Route To | Why |
|---|---|---|
| Company name present ("warm paths for Anthropic") | network-intelligence | Single-company deep analysis |
| No company name ("warm paths", "find connections") | job-search-scheduler (enrich) | Batch enrichment on latest scan |
| "who do I know at [Co]" | network-intelligence | Explicit company = single |
Rule: Company name โ network-intelligence. No company โ job-search-scheduler enrich mode.
Detection: Check the onboarding marker:
brain/identity/experience-history.md does NOT existIf the file does not exist, show the welcome and route to the onboarding wizard:
โโโ Career OS: Mission Control โโโ
Welcome to Career Intelligence. I'm your career co-pilot with persistent memory.
Before I can help, I need to learn a bit about you. This takes ~5 minutes
and only happens once. Everything you share stays private on your machine.
Say "onboard me to career intelligence" to begin, or describe your job-search goal.
If onboarding is missing:
Invoke career-intelligence-onboarding immediately. Do not ask questions yourself โ the onboarding wizard owns the interview.
Git setup: After any onboarding completes, prompt:
Your context is saved. Want to back it up to GitHub?
Say "backup" to set up git + GitHub remote โ takes 2 minutes and protects everything.
Before rendering the dashboard, check for pending skill flags:
~/.career-os-state/pending-organize exists:
โโโ Career OS: New Feature Available โโโ
The organize skill can index your stories โ adding frontmatter,
cross-references, and a searchable story index.
Say "organize" to run it now, or dismiss to skip.
When brain/identity/experience-history.md exists, show the dashboard.
Triggers: "beta metrics", "funnel", "nsm", "local telemetry"
Render the local XOS-98 telemetry dashboard by invoking the deterministic aggregator:
bun "$CLAUDE_PLUGIN_ROOT/src/telemetry/report.ts"
The aggregator reads the local events JSONL path resolved by defaultEventsPath
(CAREER_OS_EVENTS_LOG โ XOS_EVENTS_LOG โ ~/.career-os-events.jsonl), skips
malformed rows, and prints the funnel, NSM, cohort, and recent-activity view.
Mission Control must not parse the JSONL or compute conversion, drop-off, cohort,
or NSM values itself. Treat the rendered NSM as an ESTIMATE because active time is
bucketed.
NEXT_SESSION_HANDOFF.md for previous session contextgit log --oneline -5 for (handoff updated externally) โ if found,
read current handoff for another agent's updatesgh issue list --repo $CAREER_GITHUB_REPO --state open --json number,title,labels) and job-pipeline.json + job-pipeline-match-tracker.json for dashboard dataprofile_change_logged event, if any. Summarize the 30d before/after inbound impact (XOS-94) and render a single plain-text profile-impact line in the dashboard header area.emitDashboardViewed({ has_career_data, has_brand_data }) through the existing XOS-98 gated local JSONL telemetry path.For each entry in job-pipeline.json โ stage_data[], compute time-in-stage and flag:
| Stage | Stale After | Action |
|---|---|---|
| Applied, awaiting | 14 days | Flag + draft follow-up (route to outreach-composer) |
| Advancing (screen+) | 10 days since last activity | Flag + suggest ping recruiter |
| References in progress | 7 days | Flag |
| Ready to apply | 7 days | Flag + prompt to submit or drop |
If brain/config/pipeline-health.json exists, use its thresholds.
Self-evolution (P14, EXP-002): Track outcomes for threshold calibration.
After 10 outcomes โ first calibration (ยฑ1 day/cycle). Log to
brain/experiments/EXP-002-log.md.
โโโ Career OS โ [Day], [Date] โโโ
โ ๏ธ STALE ALERTS (only if any)
- {Company} ({Stage}, {N}d) โ {action}
TODAY'S PRIORITIES
๐ด P0: [Description] โ "[prompt]"
๐ด P0: [Description] โ "[prompt]"
๐ก P1: [Description] โ "[prompt]"
โณ Blocked ({N}): [summary]
PIPELINE
โ Company Stage โโโโโโ METRICS
โ Company Stage โโโโ Applications: N
โ Company Stage โโ Response rate: N%
Interviews: N
Active contacts: N
LOCAL BETA METRICS
[Render compact output from `bun "$CLAUDE_PLUGIN_ROOT/src/telemetry/report.ts"`]
NSM line must retain the ESTIMATE label from the aggregator.
WARM CONTACTS โ ACTION NEEDED (only if any due)
| Contact | Company | Status | Next Action |
COMING UP
[Date] โ [Event/deadline]
CAREER BRAIN
๐ข Stories: N ๐ข/๐ด People: N ๐ข/๐ก Index: status
QUICK ACTIONS
โโ Status โโ
โ "pipeline" Active: interviews, referrals, apply queue
โ "referrals" Referral status + overdue alerts
โ "dashboard" Scored roles ready to apply (โฅ80%)
โ "beta metrics" Local funnel, estimated NSM, cohorts
โโ Search & Score โโ
โ "scan for jobs" Find new roles with warm-path detection
โ "score the latest scan" Score new roles (Opus)
โโ Apply โโ
โ "customize resume for [Co]" Tailor resume
โ "cover letter for [Co]" Generate cover letter
โ "answer questions for #N" Portal question answers
โ "I applied to [Co]" Record application
โโ Track โโ
โ "heard back from [Co]" Process a response
โ "sync pipeline" Reconcile all files
โโ Prep & Network โโ
โ "prep me for [Co]" Interview prep
โ "follow up with [Name]" Time-calibrated follow-up
โ "who do I know at [Co]" Find warm intros
โโ Execute โโ
โ "cc" / "cruise control" Batch-execute task list
โ "mission control" Refresh dashboard
| Section | Source | Read Pattern |
|---|---|---|
| Priorities | GitHub Issues $CAREER_GITHUB_REPO | gh issue list --state open filtered by tier:p1/tier:p2, sort by tier ascending |
| Pipeline | career-intelligence/projects/job-search/job-pipeline.json โ stage_data[] | Active/advancing entries |
| Metrics | career-intelligence/projects/job-search/job-pipeline-match-tracker.json | Count by status field (APPLIED, REJECTED, INTERVIEWING, OFFERED) |
| Warm Contacts | people/*.json + people-followup-query.py | Run: python3 ~/.career-os-state/scripts/people-followup-query.py --people-dir $CAREER_HOME/network/people --days 7 --format json โ returns contacts with follow_up โค today+7d. Falls back to .md frontmatter for unmigrated files. NEVER hand-scan people files. |
| Company Action Gate | company-flags.json + company-flags-filter/HOW.py | Before surfacing ANY action for a named company, run: python3 $(ls -v ~/.claude/plugins/cache/xos/career-intelligence/*/rules/company-flags-filter/HOW.py 2>/dev/null | tail -1) '{"company":"<name>","action":"<apply|follow_up|referral>","flags_file":"$CAREER_HOME/career-intelligence/projects/job-search/company-flags.json"}' โ exit 1=BLOCK (suppress), exit 2=WARN (surface referral status instead). A 92% score on a deprioritized company must never surface as an action. Gate FAIL-HARDs (BLOCK) if flags_file path is given but the file is missing (XOS-27) โ never silent-pass on absent safety config. |
| Coming Up | GitHub Issues + job-pipeline.json โ pending_referrals[].follow_up_date | Extract due dates from issue body / due:* labels + referral follow-up dates |
| Career Brain | stories/**/*.md (recursive), people/*.json | Count .md files recursively under stories/ (stories are organized into category subdirs like stories/google/, stories/independent/). Exclude STORY_INDEX.md and README.md. For people/, count *.json files (migrated as of v0.37.0); fall back to *.md count if no JSON found. |
| Stale Alerts | job-pipeline.json โ stage_data[] | Compute days-in-stage from stage_detail date or tracker updated_at |
| Local Beta Metrics | src/telemetry/report.ts reading the local events JSONL | Run bun "$CLAUDE_PLUGIN_ROOT/src/telemetry/report.ts" and render the output; do not duplicate math in this skill |
| Profile Impact | Local telemetry JSONL | Latest profile_change_logged event + src/telemetry/profile-inbound-impact.ts summary over content_to_dm_tracked events |
Story count command (authoritative):
find -H brain/stories -type f -name "*.md" \
-not -name "STORY_INDEX.md" -not -name "README.md" 2>/dev/null | wc -l
Note: -H is required because brain/stories/ may be a symlink to $CAREER_HOME/brain/stories/ after the Brain-layer migration. Without -H, macOS find returns 0 results on a symlinked start directory.
Do NOT use ls brain/stories/*.md โ it misses subdirectory contents
and will undercount when stories are organized into categories (P9 coherence
issue fixed by WO-049).
| Condition | Action |
|---|---|
| stories/ has .md files (recursive) AND no STORY_INDEX.md | Suggest "organize" |
| STORY_INDEX.md exists but older than newest story (recursive) | Suggest "organize" (incremental) |
| STORY_INDEX.md exists and current | Show memory health counts |
| stories/ empty (recursive) or doesn't exist | Skip memory health section |
"Recursive" means: apply the same find command as the Story count command above โ include all subdirectories, exclude STORY_INDEX.md and README.md.
Include in dashboard header area:
SCAN STATUS
โ
15 roles found, 6 warm paths, 3 second-degree intros (enriched)
or
โ
15 roles found (6 AM safety net) โ say "enrich warm paths" for referral detection
or
โ ๏ธ No scan today โ say "scan for jobs"
Mission Control makes the brand โ career flywheel visible once per week using the same local attribution events surfaced by Pipeline View (XOS-102).
Read the local telemetry log via the shared weekly summary helper
src/telemetry/weekly-inbound.ts, which aggregates content_to_dm_tracked
events from $CAREER_OS_EVENTS_LOG / $XOS_EVENTS_LOG, defaulting to
~/.career-os-events.jsonl. One content_to_dm_tracked event equals one
attributed inbound recruiter DM.
Behavior:
week_of as the Monday date for the current week, formatted
YYYY-MM-DD.0, omit the card.insight_card_viewed event already exists for
insight_kind: "weekly_inbound" and this week_of, omit the card.emitInsightCardViewed({ insight_kind: "weekly_inbound", week_of }).emitInsightActedOn({ insight_kind: "weekly_inbound", action, week_of }).Plain-text card:
โโโ Weekly Insight โโโ
๐ This week: {N} recruiter DMs attributed to your posts
Top posts:
- {post_id}: {count} DMs
โ "campaign engine" Turn the top post into your next content campaign
โ "post more on {topic}" Double down on the topic that drove inbound
If only one top post exists, show one bullet. If several posts tie, show up to three. Keep the output plain-text-friendly: no markdown tables, no pipe characters, no nested cards.
Action telemetry:
action: "open_campaign_engine"action: "post_more_on_topic"Mission Control closes the feedback loop on profile optimization by tying a logged profile change to inbound recruiter DM volume over the surrounding 30d windows.
When the user says they updated or rewrote a profile section, log the change with the local telemetry helper:
emitProfileChangeLogged({ section, note })
Use section: "headline", "summary", or "experience" when known. Keep
note short and optional, such as "operator-led headline rewrite". The helper
uses the existing XOS-98 gated local JSONL emitter. It adds no transport and
makes no outbound calls.
To surface impact, read the latest profile_change_logged event from the local
telemetry log, then call summarizeProfileChangeImpact(change.ts) from
src/telemetry/profile-inbound-impact.ts. The helper counts
content_to_dm_tracked events in the 30 days before the change and the 30 days
after the change, then returns inboundRateChange = afterCount - beforeCount.
Plain-text line:
Profile impact: inbound-rate-change {+/-N} recruiter DMs after your {section} change
30d before: {beforeCount} DMs
30d after: {afterCount} DMs
Omit this line when there is no logged profile change. If the after window is still in progress, show the count accumulated so far. Keep this as text in the dashboard header area; do not add a nested card or table.
Mission Control must make the career + brand fusion loop visible in one 6-second scan. Show both halves every time the returning-user dashboard renders: career execution on one side, brand/inbound compounding on the other. If one side has no data, keep the side visible and show a gentle no-data line instead of omitting it.
Career data:
career-intelligence/projects/job-search/job-pipeline.json using the
existing Pipeline data-source pattern: stage_data[].has_career_data is true when stage_data[] has at least one entry.stage_data[]
(exclude terminal states: rejected, dropped, archived, closed, dead, deprioritized).stage_data[] entries whose stage/status/stage detail is
screening, recruiter screen, hiring manager screen, technical screen, onsite,
interview, references, or offer-process.pending_referrals[] from job-pipeline.json plus the existing
people-followup-query.py --people-dir $CAREER_HOME/network/people --days 7 --format json
pattern. Never hand-scan people files for this count.Brand/inbound data:
summarizeWeeklyContentInbound from src/telemetry/weekly-inbound.ts.profile_change_logged
event exists, call summarizeProfileChangeImpact(change.ts) from
src/telemetry/profile-inbound-impact.ts.content_to_dm_tracked records in
the same telemetry JSONL used by the existing helpers. Do not recompute weekly
groupings; the weekly count and top posts come from
summarizeWeeklyContentInbound.has_brand_data is true when at least one content_to_dm_tracked event
exists.Telemetry:
emitDashboardViewed({ has_career_data, has_brand_data }).Plain-text dashboard block:
โโโ Career + Brand Health โโโ
CAREER
Active applications: {active_applications}
Interviews: {interviews}
Warm intros available: {warm_intros_available}
Status: {if no career data, "no data yet - add an application or referral"}
BRAND / INBOUND
Weekly inbound DMs from posts: {weekly_inbound_total}
Post -> DM attributions: {post_dm_attribution_count}
Profile impact: {+/-N} DMs after {section} change
Status: {if no brand data, "no data yet - track a post-attributed inbound DM"}
Rendering rules:
Profile impact: no recent profile change logged rather than omitting the
line.After dashboard renders, check these in priority order and offer ONE prompt:
Priority order:
๐ก [One question with brief WHY] (skip to dismiss)
If user answers โ route to appropriate skill โ ask next โ repeat until user says skip/q/done or changes topic.
NEXT_SESSION_HANDOFF.md โ the previous
session's narrative (what happened, what's in-flight, decisions made)git log --oneline -5 for
(handoff updated externally) tag โ if found, read current handoffNEXT_SESSION_HANDOFF.md with this
session's narrative. Update continuously throughout the session.When ~/.career-os-state/version does NOT match plugin version:
brain/stories/retro-v{new-version}.mdOn FIRST session after plugin install (no version file yet):
After onboarding completes:
You're all set up! How was the install experience?
Any steps that were confusing or could be smoother?
Save to brain/stories/retro-v{version}-first-install.md.
Set initial version in ~/.career-os-state/version.
Origin: 2026-05-05 gate failure โ Mission Control suggested "verify IC title with Paul" without reading paul-hessey.md, where the interaction log showed this had already been confirmed via email 2026-05-04. The action was already done; the suggestion created false noise and eroded trust.
Rule: Before surfacing ANY action suggestion for a named contact โ follow-up, ping, verify, outreach, intro request, thank-you, or any other contact-directed action โ Mission Control MUST read the contact's people file first.
For each named contact about to appear in any action suggestion:
people_file = network/people/<slug>.md
If the file does not exist: emit โ ๏ธ No people file for {Name} โ action suggestion suppressed. Say "add {Name} to network" to create one. Do NOT suggest the action.
If the file exists: read it completely before proceeding to step 2.
Read the # Interaction Log section of the people file. For the proposed action, ask:
"Is there evidence in the interaction log that this action was already taken, or that the information it would surface is already known?"
| Log evidence | Verdict |
|---|---|
| Exact same action appears in log (e.g., "Citizenship confirmed") | SUPPRESS โ do not suggest; it is already done |
| Related action appears (e.g., email response received after proposed follow-up) | SUPPRESS or DOWNGRADE โ check if the follow-up is still relevant |
| No log entry at all for this action | ALLOW โ safe to suggest |
Read last_contact: from the frontmatter.
last_contact within 7 days: contact is warm. Suggest actions conservatively โ don't pile on.last_contact between 7โ14 days: note the gap inline: (last contact {N}d ago).last_contact > 14 days or missing: flag as stale before suggesting action.When a contact action passes all three steps, surface it as:
WARM CONTACTS โ ACTION NEEDED
{Name} ({Company}) โ {action} ยท last contact {N}d ago
When suppressed by log check, emit nothing (silent suppression โ the human doesn't need to know about actions that are already done).
Litmus test: "Did I read the full people file for {Name} before suggesting this action? Did I check the interaction log for evidence this is already done?" If either answer is no โ gate failed.
brain/ file โ route to owning skill