pipeline-hygiene-audit
Skill5 starsChanged 4 months agoMITView on GitHub
Finds the dead deals, stale stages and missing dates that make a CRM pipeline unreliable, and separates real pipeline from wishful pipeline. Use when a forecast or board update is due, or when the number in the CRM has stopped resembling reality.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- elijeangilles/revops-skillstheir own version, first seen Jun 2, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- elijeangilles/revops-skills★ 9same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository30 other files it gives its agents.
- AGENTS.mdAGENTS.md
- ad-to-landing-promise-matchSkill
- b2b-audience-signal-auditSkill
- cold-outbound-sequence-reviewSkill
- competitor-lead-angle-reviewSkill
- crm-lead-source-quality-auditSkill
- discovery-call-gap-analysisSkill
- disqualification-reason-minerSkill
- email-nurture-sequence-reviewSkill
- form-friction-finderSkill
- ghosted-after-the-demoSkill
- google-ads-lead-quality-auditSkill
- icp-reality-checkSkill
- landing-page-qualification-reviewSkill
- lead-magnet-offer-fitSkill
- lead-scoring-sanity-checkSkill
- lead-to-opportunity-drop-off-reviewSkill
- linkedin-ads-lead-quality-auditSkill
- meeting-prep-five-minutesSkill
- meta-ads-lead-quality-auditSkill
- objection-cheat-sheetSkill
- proposal-win-loss-reviewSkill
- sales-follow-up-speed-auditSkill
- scale-readiness-check-lead-genSkill
- sdr-handoff-brief-generatorSkill
- spam-folder-checkSkill
- weekly-b2b-lead-gen-readoutSkill
- what-to-say-to-this-companySkill
- won-deal-to-case-studySkill
- write-the-follow-upSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 12 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈12
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Run this before the forecast, not after it goes wrong.
- Very specific: line 31
Run ../scripts/pipeline_hygiene.py over the export.
- Distinctive: line 34
Check the close dates. Dates in the past, dates that have been pushed more than twice, and dates that cluster suspiciously on the last day …
- Distinctive: line 35
Check stage definitions against stage behaviour.
- Distinctive: line 41
No activity plus stage age beyond the cycle equals a close-lost candidate, whatever the owner says about it.
- Distinctive: line 42
Never delete or alter deals.
- Distinctive: line 44
If stage definitions are missing or contradictory, fix that before trusting any conversion rate calculated from them.
- Distinctive: line 45
Do not report a forecast number this skill has not been given the inputs to support.
- Distinctive: line 85
Do not modify CRM records.
- Distinctive: line 86
Do not report a hygiene-adjusted number as a forecast.
- Distinctive: line 87
Do not name individual owners in a way that turns a data exercise into a performance review.
- Distinctive: line 88
Do not assume a missing field means a bad deal.
- Distinctive: line 89
Do not treat a single quarter as the sales cycle if the data covers less than two full cycles.
≈ “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.
Files
Pipeline Hygiene Audit
Use this skill when
The pipeline number is quoted in meetings and nobody quite believes it.
Pipeline decays quietly. A deal that was real in March sits in Proposal in August because nobody wants to be the person who closes it as lost. Multiply that by a year and the total at the top of the report stops meaning anything, which is usually discovered during a forecast conversation with someone senior.
Run this before the forecast, not after it goes wrong.
Required input
- A deal export with, at minimum: deal name or id, stage, value, created date, and last activity date.
- What the stages are supposed to mean.
Better with:
- expected close date, and whether it has been changed before
- owner
- source
- historical conversion rate between stages
- typical sales cycle length
Analysis workflow
- Run
../scripts/pipeline_hygiene.pyover the export. Counting stale deals, measuring stage ageing and comparing close dates against the sales cycle are arithmetic, and doing it by eye across a few hundred deals produces numbers that feel right and are not. - Age every deal against the typical sales cycle for its stage. A deal sitting in one stage for several times the normal duration has usually stopped moving rather than slowed down. Pick the multiplier from this store's own cycle rather than a rule of thumb.
- Find deals with no activity inside a reasonable window. No activity is the single strongest predictor of a deal that will never close, and it is usually visible months before anyone acts on it.
- Check the close dates. Dates in the past, dates that have been pushed more than twice, and dates that cluster suspiciously on the last day of a quarter each tell a different story about how the forecast is being constructed.
- Check stage definitions against stage behaviour. If deals routinely skip a stage or sit in one for a day, the stage is not doing any work and the pipeline reporting inherits that.
- Split the pipeline into what is defensible and what is decoration, and give both numbers. The gap between them is the actual finding.
- Produce a specific list to review, deal by deal, rather than a recommendation to "clean up the CRM", which never happens.
Decision rules
- No activity plus stage age beyond the cycle equals a close-lost candidate, whatever the owner says about it.
- Never delete or alter deals. This produces a review list; a human decides.
- A pushed close date is information, not a problem in itself. Three pushes on the same deal is a pattern.
- If stage definitions are missing or contradictory, fix that before trusting any conversion rate calculated from them.
- Do not report a forecast number this skill has not been given the inputs to support. Quote the defensible subset instead.
Output format
Two numbers
Total pipeline as reported, and defensible pipeline after removing stale and inactive deals. State what was removed and why.
Hygiene table
| Issue | Deals | Value | What it means |
| No activity beyond window | |||
| Stage age beyond cycle | |||
| Close date in the past | |||
| Close date pushed 3+ times | |||
| Missing required field |
Review list
The specific deals to look at, ordered by value, with a recommended action each: close lost, re-engage, re-date, or leave.
Stage health
Where deals actually get stuck, and whether any stage is doing no work.
What to fix in the process
The one or two changes that would stop this recurring, such as an activity-based ageing rule.
Practical example
An export of 340 open deals showing 4.1M in pipeline.
The script finds 96 deals with no activity in over 90 days, worth 1.3M, and a further 40 whose close date has passed. The stage ageing shows Proposal holding deals for a median of 71 days against a stated 21-day cycle.
The defensible number is 2.4M, not 4.1M, and that gap is the whole finding. The review list leads with the eleven highest-value inactive deals rather than asking anyone to review all 96. The process fix is a rule that flags a deal at 45 days without activity, which would have caught most of these while they were still recoverable.
Guardrails
- Do not modify CRM records.
- Do not report a hygiene-adjusted number as a forecast. It is a cleaner input to one.
- Do not name individual owners in a way that turns a data exercise into a performance review.
- Do not assume a missing field means a bad deal. Missing data is missing data.
- Do not treat a single quarter as the sales cycle if the data covers less than two full cycles.
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
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.

