pipeline-view

name	pipeline-view
description	Shows the live job pipeline — 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".
triggers	["pipeline","pv","show pipeline","active pipeline","referrals","show referrals","referral status","apply queue","show queue","pipeline summary","who's advancing","what's active","pipeline view"]

Pipeline View — Career OS Skill

Purpose

Human-readable view of the job pipeline, rendered on demand from JSON sources. This skill has one job: run pipeline-view.py with the right flags and show the output. No state mutations. No writes.

Output Format

Always start with:

━━━ Career OS: Pipeline View ━━━

How to Invoke

User says	Section shown
pipeline / show pipeline	All sections
active pipeline / who's advancing	Active/advancing only
referrals / show referrals / referral status	Pending referrals with overdue alerts
applied / what have I applied to	Applied roles awaiting response
apply queue / show queue	Scored roles not yet applied, ranked by score
pipeline summary	Summary metrics only

---

BEHAVIOR

Step 1: Map intent to --section flag

Trigger pattern	Flag
referral(s)	--section referrals
active / advancing / interviews	--section active
applied / awaiting response	--section applied
queue / apply queue	--section queue
summary / metrics	--section summary
anything else / no qualifier	--section all

Step 2: Run the script

python3 ~/.career-os-state/scripts/pipeline-view.py \
  --career-home $CAREER_HOME \
  --section <section>

$CAREER_HOME must be set to the customer's workspace root (typically ~/career-os or wherever they keep their career data). Run career-intelligence-onboarding if unsure.

Step 3: Print the output verbatim

Render the script's stdout directly — do not paraphrase, summarize, or reformat it. The script owns the display logic. If content-attributed inbounds exist, append the XOS-102 block below after the script output.

If the script errors (file not found, JSON parse error), surface the error message and suggest sync pipeline to diagnose data health.

Step 4: Content-attributed inbound DMs (XOS-102)

The DM count is per-DM, NOT per-contact — one recruiter can send multiple post-attributed DMs, so counting contacts under-counts. The per-DM unit of record is the content_to_dm_tracked telemetry event (one event = one attributed DM).

Count = the number of content_to_dm_tracked events in the telemetry log ($CAREER_OS_EVENTS_LOG / $XOS_EVENTS_LOG, default ~/.career-os-events.jsonl). Each event carries post_id + dm_source (+ contact_slug). List one line per event. Append a plain-text-friendly block:

━━━ Content → Inbound DMs ━━━
{N} inbound DMs attributed to your posts.
- inbound from {contact_slug} via post {post_id}
  source: {dm_source}

Fallback when the telemetry log is absent (e.g. XOS_98_TELEMETRY was off): read $CAREER_HOME/network/people/*.md frontmatter, count contacts with source: post + non-empty post_id, and label the number a lower bound ("≥{M} (contact-level; enable XOS_98_TELEMETRY for per-DM counts)") since it collapses multiple DMs per contact.

If there are zero post-attributed inbounds (no events AND no flagged contacts), omit the block unless the user asked specifically for content attribution.

---

ROUTING

This skill is read-only. For any mutations, route to the appropriate skill:

User follow-up	Route to
"I applied to #N" / mark as applied	apply-tracker
"follow up with [Name]"	outreach-composer
"prep me for [Company]"	interview-prep
"update BuildOps status"	apply-tracker
"sync pipeline" / "something looks wrong"	pipeline-sync
"score the queue"	job-match-scorer

---

FOOTER (always append after output)

━━━ Actions ━━━
→ "applied to #N"              — record application
→ "follow up with [Name]"      — draft follow-up message
→ "pipeline summary"           — metrics only
→ "referrals"                  — referral status + overdue alerts
→ "sync pipeline"              — diagnose data health