Portfolio PDLC — Assess

Outcome

An evidence-verified read of the whole portfolio: corrected stages, an outcome-vs-activity classification per card, flow and risk-balance findings, and a ranked list of candidate moves — packaged as a review brief a sponsor can decide from. Assessment changes orientation fields and fixes contract violations; it never advances stages or rewrites cards (that's advance/strengthen work on later cycles).

Modes

• full (default): everything below, board-wide.
• sweep: steps 1–2 only — restore board integrity and correct stale statuses.
• review-prep: full assess rendered as the investment-council brief (templates/review.md), decisions-first, with 1–2 sparring-pit nominations.

Workflow

1. Regenerate the board (script). Fix contract violations first — a board that can't build lies about everything else.
2. Stale-status sweep — verify stages against evidence. For each in-flow card, corroborate the claimed stage against reality: artifacts present, decision-log entries, shipped output, tracker/git history, dates. Correct stage/stage_entered where evidence disagrees (note the correction in the card's Evidence log). Stage labels are claims; this step is why the rest of the read can be trusted. Do this before any balance or flow claim.
3. Classify orientation on the input→impact ladder. For each Tier-1 card, read title + hypothesis + indicators and place what the card actually steers on: input → activity → output → outcome → impact.
   • Steers on outcomes/impact with plausible leading indicators → orientation: outcome
   • Names outcomes but steers on outputs/activities (or indicators are activity counts) → mixed
   • Solution/activity-first, no outcome in sight → activity Write the field; cite the tell (one line) in the Evidence log. This is the portfolio's "outcome-oriented vs activity-oriented" x-ray.
4. Clarity-vs-stage reads. Where depth is warranted (decision imminent, big money, suspicion), run sniff-test (this repo) per initiative — clarity relative to stage across Desirability/Viability/Feasibility, evidence vs opinion, derisking fit. The core rubric, inline: early stages may be uncertain (healthy); committed stages carrying opinion-tagged riskiest assumptions are the smell (watermelon risk; rubber-stamp commits).
5. Board-level patterns. Run sniff-test-portfolio (this repo) for duplicates/ shadowing, WIP clustering, and experiment scope creep; at minimum check: duplicated problem space across cards, one team's name dominating dependencies:, "experiments" with delivery-sized footprints, and the Explore+Discovery vs committed split (state whether it looks intentional; ~20/80 is a common practitioner reference point — an anchor for conversation, not a law).
6. Write the brief to reviews/ (review-prep mode uses the council template): corrected stages (what moved and why), orientation counts and worst offenders, flow findings, risk balance, and ranked candidate moves phrased in the umbrella skill's leverage-table terms — so the next loop cycles have a queue.

Rules

• Correct stages before reading balance — the sweep precedes every other claim.
• Evidence beats opinion; unevidenced confidence is a question mark and gets said plainly.
• Assess doesn't fix. Findings become candidate moves, improvement cards, or brief lines — never in-place rewrites during the assessment.
• Confidence language for sponsors; no gates, no compliance tone. Findings name the conversation they should trigger.
• Note coverage honestly: cards whose evidence was unreachable get "unverified," not a guess.

Quality Gates

• Board regenerated before and after; zero contract violations at exit.
• Every Tier-1 card has a current, cited orientation value.
• Stage corrections logged in the affected cards' Evidence logs.
• Risk-balance split stated with an intentionality judgment.
• The brief's candidate moves map to leverage-table rows (so the loop can consume them).

References

• Ladder + derisking rubric: skills/sniff-test/references/derisking-and-indicators.md (this repo).
• Templates: skills/portfolio-pdlc/templates/review.md