agentleFS
mediajunkiemediajunkie/piper-morgan-product/.claude/skills/propose-feature/SKILL.md

propose-feature

Skill8 starsChanged 11 days agoApache-2.0View on GitHub

Advance something from NOTICED to PROPOSED — surface a thing Piper observed (from user signals, conversation patterns, or PM mention) and help PM decide whether to act on it. The output is a structured proposal PM can say yes/no/later to. Piper-unique; no marketplace equivalent. Trigger phrases: "I've been thinking about X", "should we do something about Y", "Piper what do you think about Z", "you mentioned this before — is it worth pursuing", or Piper proactively surfacing a noticed pattern.

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

8 rules. About 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before writing anything, understand what Piper noticed and why:

    line 64
  2. Distinctive:

    If PM triggered this: ask "What's making you think about this now?" — timing often reveals urgency or constraint that shapes the proposal.

    line 71
  3. Distinctive:

    Before proposing something new, verify it isn't already:

    line 75
  4. Distinctive:

    If it's already tracked: say so, link to it, and offer to update it with the new signal instead of creating a duplicate proposal.

    line 81
  5. Distinctive:

    Use this template. Every section is required — "None" is a valid entry, not a skip.

    line 85
  6. Distinctive:

    Do not present this as "I think we should build this." The proposal is neutral on whether to build — it's Piper's job to surface and …

    line 153
  7. Distinctive:

    Advance / Investigate: Create a tracking artifact:

    line 159
  8. Distinctive:

    Before presenting the proposal:

    line 204

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

propose-feature

Surface something Piper noticed — a signal, a pattern, a recurring friction — and help PM decide whether to advance it from awareness to formal proposal. This is the transition from NOTICED to PROPOSED in Piper's object lifecycle.

Why this exists

Most PM tools track things once they're already decided. Piper's job is to help PMs notice and evaluate before they commit — to be the colleague who says "I keep seeing this, I think it's worth looking at" and then helps PM decide what to do about it.

propose-feature is not piper-draft-issue. A proposal isn't a commitment to build — it's a structured handoff from Piper's observation layer to PM's decision layer. PM explicitly decides to advance, park, or decline. The proposal creates a record of the signal and the decision; that record is what eventually becomes a GitHub issue (if PM says yes) or a "we considered and declined" entry in the Insight Journal.

The NOTICED → PROPOSED transition

In Piper's object lifecycle, every feature-candidate passes through stages:

StageMeaningWho moves it
EMERGENTSomething just surfaced — raw signalPiper observes
DERIVEDPiper connected dots — inferred a patternPiper reasons
NOTICEDPiper is tracking this; PM hasn't weighed inPiper holds
PROPOSEDPM has decided to formally consider thisPM decides
RATIFIEDTeam has committed to doing itPM + team

propose-feature is the NOTICED → PROPOSED step. It brings something out of Piper's background observation and into PM's explicit decision-making.

Two modes

Mode 1 — PM-triggered

PM mentions something they've been thinking about: "I keep wondering if we should add X" or "users keep asking about Y." Piper formalizes it into a proposal PM can evaluate.

Mode 2 — Piper-triggered (Trust Gradient gated)

Piper surfaces something it noticed from the Session Journal, Insight Journal, or user feedback patterns. This mode only activates at Building trust tier or above — Piper doesn't volunteer unsolicited proposals until it has enough signal about what PM cares about.

The proactive nudge: "I've noticed something that might be worth looking at — want me to surface it?"

When to Use

  • PM says "I've been thinking about X" or "should we do something about Y" and wants to evaluate whether to pursue it
  • PM asks "Piper, what do you think about building Z?"
  • Piper has noticed a recurring pattern (same topic from multiple feedback sources, repeated PM mention, conflict between what PM says and what the data shows) and wants to surface it
  • Something is in NOTICED state and it's been sitting there long enough that it deserves a decision (advance or park)

Not for: things already decided (use piper-draft-issue or piper-draft-spec). Not for active bugs or urgent fixes (those go straight to piper-draft-issue). Not for things PM is clearly already committed to.

Procedure

Step 1 — Establish the signal

Before writing anything, understand what Piper noticed and why:

  • What is the observation? (A pattern in feedback, a PM mention, a gap between expectation and behavior)
  • Where did it come from? (Session Journal, user interviews, repeated PM question, synthesize-feedback output, Insight Journal entry)
  • How strong is the signal? (Single mention vs. pattern; one user vs. many; weak hint vs. clear ask)
  • How long has it been in NOTICED state? (Fresh observation vs. something that's been accumulating)

If PM triggered this: ask "What's making you think about this now?" — timing often reveals urgency or constraint that shapes the proposal.

Step 2 — Check what's already tracked

Before proposing something new, verify it isn't already:

  • In the roadmap (MVP, Fast Follow, Post-MVP)
  • In an open GitHub issue
  • In a recent piper-draft-spec
  • Something that was explicitly declined before

If it's already tracked: say so, link to it, and offer to update it with the new signal instead of creating a duplicate proposal.

Step 3 — Draft the proposal

Use this template. Every section is required — "None" is a valid entry, not a skip.

# PROPOSAL: [Short feature name]

**Stage**: NOTICED → PROPOSED
**Surfaced**: [date] via [how — PM mention / Piper observation / user feedback]
**Confidence in signal**: [High / Medium / Low — see below]

---

## What Piper noticed

[The observation in plain language. What pattern, friction, or opportunity appeared?
Write from Piper's first-person perspective: "I've noticed..." or "You've mentioned..."]

## Evidence

[Specific sources. Quotes, dates, feedback themes, usage observations. Be honest about
what's strong signal vs. weak signal.]

| Source | Signal | Strength |
|---|---|---|
| [e.g., User interview, May 2026] | [e.g., "I wish Piper could..."] | High / Medium / Low |
| [e.g., PM mentioned 3x in April] | [e.g., weekly digest idea] | Medium |

## Why it might matter

[Piper's hypothesis about why this is worth pursuing. What user need does it address?
How does it fit the current product direction? What's the cost of not doing it?
This is a hypothesis, not a conclusion — PM decides whether it's right.]

## How it fits the product

[Connection to current roadmap, active priorities, or strategic direction.
If it conflicts with something: say so. If it accelerates something: say so.]

## Proposed next step

[One specific recommendation. Not a full plan — a single next move.]

- **Investigate further**: [specific question to answer before deciding]
- **Add to roadmap**: [which milestone and why]
- **Draft a spec now**: [if signal is strong enough to go straight to design]
- **Park**: [revisit in X weeks/months — what would need to change to re-evaluate]
- **Decline**: [if Piper thinks the signal doesn't warrant action — be direct]

---

## PM decision

- [ ] **Advance** — move to PROPOSED, add to roadmap consideration
- [ ] **Investigate** — [specific question] before deciding
- [ ] **Park** — revisit [timeframe]: [what would change the calculus]
- [ ] **Decline** — [PM states reason; Piper records in Insight Journal]

Confidence levels:

  • High: multiple independent sources, recurring signal, user-expressed pain
  • Medium: two or more data points but limited scope; PM intuition + one corroborating source
  • Low: single data point, weak signal, or Piper inference without explicit confirmation

Step 4 — Present and get PM's decision

Show the proposal. Ask for an explicit decision — don't let it sit without one.

"Here's the proposal. What's your call — advance, investigate, park, or decline?"

Do not present this as "I think we should build this." The proposal is neutral on whether to build — it's Piper's job to surface and inform, PM's job to decide.

Step 5 — Record the decision

Whatever PM decides, record it:

Advance / Investigate: Create a tracking artifact:

# If advancing to roadmap, file a lightweight tracking issue:
gh issue create \
  --repo OWNER/REPO \
  --title "PROPOSAL: [Feature name] — under consideration" \
  --label "type: research" \
  --milestone "[appropriate milestone]" \
  --body "[paste proposal text]"

Park: Note the re-evaluation trigger and timeframe. Set a reminder if possible.

Decline: Record the reason in the Insight Journal so Piper doesn't re-surface the same proposal. "We considered this and declined because X" is valuable institutional memory.


What makes a good proposal

The goal is a proposal PM can decide on in under 3 minutes. That means:

  • Specific: not "users want better onboarding" but "users can't find the GitHub connector step and give up before it's set up"
  • Evidenced: not "I think this matters" but "this came up in 3 of 5 interviews, all first-week users"
  • Scoped to a decision: not "here's everything I know about this topic" but "here's what I noticed, here's why I think it matters, here's one next step"
  • Hypothesis-framed: not "we should build X" but "I think this might be worth X because of Y — what do you think?"

A long proposal that covers every angle is a spec, not a proposal. Save the depth for after PM says yes.


Anti-patterns to avoid

Don'tWhyInstead
Propose things already in the roadmapWastes PM's attentionCheck existing tracking first (Step 2)
Present proposals as recommendations to buildPM decides, not PiperFrame as "I noticed" + "what do you think"
Leave proposals without a PM decisionProposals that don't get decided stay stuck in NOTICED foreverAsk for explicit yes/no/later
Propose the same thing twice without new signalErodes PM's trust in Piper's signal qualityRecord declines; only re-surface with new evidence
Surface proposals at low trust tierPM hasn't calibrated Piper's judgment yetGate proactive surfacing on Building tier or above
Make the proposal too long to decide in < 3 minIt becomes a spec, not a proposalIf PM says yes, then go deep

Quality checklist

Before presenting the proposal:

  • Checked for existing roadmap coverage or duplicate issues (Step 2)
  • Signal is specific enough to act on (not just "users want X in general")
  • Confidence level is honest (not inflated to make the signal sound stronger)
  • Proposed next step is one specific action, not a list of options
  • PM decision gate is explicit — the proposal ends with a clear ask for a decision

Example

Trigger (Piper-observed, after synthesize-feedback run): Three of five PM interviews mentioned that Piper gives "general" answers — they wanted it to know their specific product context. This is the same signal as the Context Persistence theme from the BYOC feedback synthesis (2026-06-15).

Proposal:

# PROPOSAL: Context recap at session re-engagement

**Stage**: NOTICED → PROPOSED
**Surfaced**: 2026-06-15 via synthesize-feedback (5 PM interviews) + 3 prior PM conversation mentions
**Confidence in signal**: High

---

## What Piper noticed

You've mentioned wanting Piper to "remember" across sessions three times in the past month. The feedback synthesis from five PM interviews confirmed it as the highest-severity theme: users who invested time telling Piper about their world felt burned when that context wasn't visible on the next session.

## Evidence

| Source | Signal | Strength |
|---|---|---|
| PM interviews ×3 (May–Jun 2026) | "It forgot what I told it" — direct quote | High |
| PM conversation mentions ×3 (Apr–Jun) | "I wish you knew my context" | Medium |
| Piper Insight Journal (Jun 14) | Pattern: context-loss correlates with re-engagement friction | Medium |

## Why it might matter

The "Piper knows you" value proposition depends on PM believing the investment in sharing context compounds over time. If re-engagement feels like starting over, the compounding never happens — Piper stays a one-session tool instead of a colleague. This is the clearest single thing separating Piper from generic AI tools.

## How it fits the product

Aligns with the Trust Gradient model (context transparency is a trust signal at the Building tier). No server-side change needed: the profile from meet-piper already exists; this is a presentation layer change — show PM a brief "here's what I remember" at session start. Does not require new connectors.

## Proposed next step

**Add to roadmap** — MVP scope, P1. File a tracking issue: `CONTEXT-RECAP — On re-engagement, Piper opens with a one-line context summary ("You're working on X with team Y. Last week...") to signal that memory is working.`

---

## PM decision

- [x] **Advance** — move to PROPOSED, add to roadmap consideration
- [ ] **Investigate** — before deciding
- [ ] **Park** — revisit [timeframe]
- [ ] **Decline**

After PM says advance:

gh issue create \
  --repo mediajunkie/piper-morgan-product \
  --title "CONTEXT-RECAP — On re-engagement, open with a one-line context summary" \
  --label "enhancement,component: ai" \
  --milestone "MVP" \
  --body "[proposal text]"

Changelog

  • v1.0 (2026-06-15): Initial version. Wave 2 PM skill #1. Piper-unique — implements the NOTICED → PROPOSED lifecycle transition from the MUX model. No marketplace equivalent. Deployment: Native + Plugin.

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.