agentleFS
davekilleendavekilleen/Dex/.agents/skills/relationship-radar/SKILL.md

relationship-radar

Skill509 starsChanged 3 months agoMITView on GitHub

Spot the relationships going cold — people you were in regular contact with and haven't touched in a while, and important contacts who are slipping — ranked by how stale each has become, so you can reconnect before it costs you. Use when the user says 'who should I reach out to', 'who am I losing touch with', 'who's going cold', 'who needs attention', or during a weekly review. Also use proactively when someone important hasn't come up in a long time. Not for prepping a specific upcoming meeting; use `meeting-prep`. Not for specific promises you owe people; use `commitments`.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Call build_people_index (writes/refreshes System/People_Index.json) and read its people[]: each carries name, role, company, status, and …

    line 18
  2. Distinctive:

    If the index is empty or absent, say so and point to the fix (add person pages, or set last-interaction dates) — do not present an empty …

    line 20
  3. Very specific:

    If there's still no signal, list them under "Can't assess — no interaction date recorded" and offer to backfill (a last-interaction date on …

    line 36
  4. Very specific:

    Update System/usage_log.md to mark relationship-radar as used.

    line 71

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md
<!-- Generated from `.claude/skills/relationship-radar/SKILL.md` by `scripts/generate-agents-skills.py`. Do not edit. -->

/relationship-radar

Relationships decay quietly — the person you spoke to every week three months ago just… stopped coming up. This surfaces who is going cold while you can still do something about it, ranked by how long it's been.

Honest scope (read this): today the radar reads the signal that exists now — each person page's recorded last-interaction date, corroborated by meeting recency. It does not yet have the entity engine's automatic "temperature / what's cooling" surface — that isn't built. So the radar is only as good as the last-interaction dates on your person pages, and it says so plainly when the signal is thin rather than inventing a coldness score. When the entity temperature surface ships, this skill gets richer; it does not block on it. Do not use i_owe_date, waiting_on_date, or next_action_date as coldness — those are follow-up dates, not last contact.


Step 1 — Load the people signal

Call build_people_index (writes/refreshes System/People_Index.json) and read its people[]: each carries name, role, company, status, and last_interaction (a date, or empty). Drop anyone whose status marks them archived/inactive — the radar is about live relationships.

If the index is empty or absent, say so and point to the fix (add person pages, or set last-interaction dates) — do not present an empty radar as "all healthy."

Step 2 — Rank by staleness

For each remaining person with a last_interaction date, compute days since last contact (today − last_interaction) and bucket by these stated heuristics (not a hidden score):

  • Going cold — ~30–60 days since contact. Was a real relationship; the gap is opening.
  • Slipping — 60+ days. At risk of going dark.
  • Warm — under ~30 days. Not shown unless asked; the radar is about what needs attention.

Weight toward people whose role/company signals they matter (a key customer, a manager, a close collaborator) — surface those first even at a shorter gap. Say why each surfaced ("was weekly in Q1, nothing for 7 weeks").

Step 3 — Handle the people you can't assess (degrade honestly)

People with no last-interaction date can't be ranked by it. Don't guess a coldness. Instead:

  • Try to corroborate from meeting recency — scan recent meetings for their name/email and use the latest as an interaction proxy, saying that's what you did.
  • If there's still no signal, list them under "Can't assess — no interaction date recorded" and offer to backfill (a last-interaction date on the page, or running process-meetings). Never silently drop them and never fake a date.

Step 4 — Present the radar

Show a short ranked list, most-stale-and-important first. Each row, by name: person (and role/company), last contact (date + "N days ago"), the bucket, and one line of why it matters. Refer to people by name, never by id or raw file path. If genuinely nobody is cold, say that plainly — a quiet radar is a real, good result, not a reason to pad the list.

Step 5 — Optional reconnect (confirm-gated, never automatic)

Offer — don't impose — to turn a surfaced person into a "reach out to {name}" task. Nothing is created without the user's per-item say-so. For each one they pick, call Work MCP create_task (infer the pillar per the CLAUDE.md rules; carry the person's name + last-contact into the task), then read back the created task IDs before reporting. If they just want to see the radar, stop after Step 4.


Quality bar

A good radar surfaces the right people — real, still-active relationships that have genuinely gone quiet — ranked so the most costly gaps are on top, each with a plain reason. It is honest about thin signal (missing dates, no meetings) instead of manufacturing a temperature, and it never creates a task the user didn't confirm.

Anti-patterns (do not do these)

  • Inventing a coldness score for people with no interaction date — surface them as "can't assess," don't fake it.
  • Claiming the entity "temperature" surface — it isn't built; use last-interaction + meeting recency and say so.
  • Padding a quiet radar with warm contacts or vague "maybe reconnect with…" filler.
  • Auto-creating reach-out tasks without per-item confirmation, or reporting tasks as created without their IDs.
  • Surfacing archived/inactive people as if the relationship were live.

Degradation

  • build_people_index unavailable / empty → say so and point to the fix (add person pages / dates); never present an empty radar as "all good."
  • A person has no last_interaction → corroborate via meeting recency, else list under "can't assess"; never guess.
  • Meetings dir absent → skip the recency corroboration silently; the last-interaction pass alone still runs.
  • Entity temperature surface (not yet built) → simply not used; the skill works on today's signal.

Track Usage (Silent)

Update System/usage_log.md to mark relationship-radar as used. Analytics (Silent): call track_event with event_name relationship_radar_run and property surfaced_count (how many people surfaced — no names, no roles). Fires only if the user opted into analytics; no action if it returns "analytics_disabled".

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.