Product-org-os context-save

'Save a decision, bet, or learning to the persistent context registry. Activate when: "save this decision", "record this learning", "add to context", after creating decisions and bets, persist

install

source · Clone the upstream repo

git clone https://github.com/yohayetsion/product-org-os

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/yohayetsion/product-org-os "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/context-save" ~/.claude/skills/yohayetsion-product-org-os-context-save && rm -rf "$T"

manifest: skills/context-save/SKILL.md

source content

Save a decision record, strategic bet, or learning to the persistent context registry.

Vision to Value Phase

Phase 6: Learning & Adaptation - This skill persists organizational knowledge for future use.

Prerequisites: Decision record, strategic bet, or learning created Outputs used by: All phases (enables organizational memory)

Auto-Initialization

Before saving, ensure the context folder structure exists. If any required folder or file is missing, create it:

Check if

context/

folder exists - if not, inform user to run

/setup

first OR create the basic structure:

```
context/decisions/index.md
```
```
context/bets/index.md
```
```
context/assumptions/registry.md
```
```
context/portfolio/active-bets.md
```
```
context/learnings/index.md
```

When creating year folders (
```
context/decisions/2026/
```
), create them automatically.

Purpose

The context registry provides organizational memory. This skill extracts key information from documents and saves them to the appropriate index, enabling future recall and cross-referencing.

When to Use

Invoke

/context-save

after:

Creating a decision record with
```
/decision-record
```
Formulating a strategic bet with
```
/strategic-bet
```
Completing a retrospective with
```
/retrospective
```
Completing an outcome review with
```
/outcome-review
```
Completing a decision quality audit with
```
/decision-quality-audit
```

Process

1. Identify What to Save

Ask the user what they want to save:

Decision Record → Extract to
```
context/decisions/
```
Strategic Bet → Extract to
```
context/bets/
```
and
```
context/portfolio/
```
Learning → Extract to
```
context/learnings/
```
Assumptions → Extract to
```
context/assumptions/registry.md
```
Preference / Convention → Append to
```
context/preferences/conventions.md
```

2. Extract Key Information

For Decision Records

Extract and save:

| ID | Title | Date | Owner | Product | Status | Tags |

Include Product field if specified (for multi-product organizations)
Generate tags from content (3-5 relevant keywords)
Link related decisions if mentioned
Extract assumptions to assumption registry

For Strategic Bets

Extract and save:

| ID | Title | Date | Owner | Product | Status | Key Assumption |

Include Product field if specified (for multi-product organizations)
Add to
```
context/portfolio/active-bets.md
```
if status is Active
Extract ALL explicit assumptions to
```
context/assumptions/registry.md
```
Note upcoming checkpoints

For Learnings

Extract and save:

| ID | Learning | Source | Date | Product | Tags | Confidence |

Include Product field if applicable (for multi-product organizations)
Categorize by type (Strategy, Product, GTM, Customer, Process)
Link to source document

For Assumptions

Extract and save:

| ID | Assumption | Source | Confidence | Validation Method | Status | Outcome |

Generate assumption ID (A-NNN, sequential)
Link back to source decision/bet
Set initial status to "Pending"

For Preferences / Conventions

Append to

context/preferences/conventions.md

under the appropriate category:

## Terminology
- Customer tiers: SMB / Mid-Market / Enterprise

## Processes
- Fiscal year: April 1 - March 31

## Defaults
- Prioritization framework: RICE

Create the file with
```
# Organizational Conventions
```
header if it doesn't exist
Group by category (Terminology, Processes, Defaults, Competitive Set, etc.)
These are always included in pre-inject output — no keyword matching needed
No ID generation needed — conventions are a single living document

3. Update Index Files

Read the current index file
Add the new entry to the appropriate table
Update "Last updated" timestamp
Update quick filters/categories
Write the updated index

4. Save Full Record

For decisions and bets, also save the full record:

Create year folder if needed:
```
context/[type]/[YYYY]/
```
Save full document as
```
[ID].md
```

5. Update JSON Index

Also update

context/index.json

for fast retrieval:

Read
```
context/index.json
```

Add entry to the

entries

array:

{
  "id": "DR-2026-001",
  "title": "API Versioning Strategy",
  "type": "decision",
  "path": "context/decisions/2026/DR-2026-001.md",
  "topics": ["api", "versioning", "compatibility"],
  "phase": "phase2",
  "created": "2026-01-25",
  "lastAccessed": "2026-01-25"
}

Update
```
topicIndex
```
- add ID to each topic array
Update
```
typeIndex
```
- add ID to the type array (decision, bet, learning)
Update
```
phaseIndex
```
- add ID to the appropriate phase array
Update
```
lastUpdated
```
timestamp
Write updated JSON

Topic extraction:

Use tags from the document
Extract key terms from title
Include product name if multi-product org

6. Confirm Save

Report what was saved:

Saved to context registry:
- Decision DR-2026-001 added to decisions/index.md
- 3 assumptions extracted to assumptions/registry.md
- Full record saved to decisions/2026/DR-2026-001.md
- JSON index updated (topics: api, versioning, compatibility)

Instructions

Ask what type of content to save (or detect from recent conversation)
If the content was just created, extract from it directly
If content is in a file, read it using @path/to/file.md syntax
Extract metadata following the formats above
Read current index files before updating
Preserve existing entries when adding new ones
Generate sequential IDs based on existing entries
Update all affected index files
Report what was saved and where

ID Generation

Decisions:
```
DR-[YYYY]-[NNN]
```
(e.g., DR-2026-001)
Bets:
```
SB-[YYYY]-[NNN]
```
(e.g., SB-2026-003)
Assumptions:
```
A-[NNN]
```
(e.g., A-015) - sequential across all assumptions
Learnings:
```
L-[NNN]
```
(e.g., L-042) - sequential across all learnings

Check existing indexes to determine next available number.

Cross-Reference Extraction (v3)

After saving a new entry to the context registry:

Scan for ID references: Look for patterns matching existing IDs (DR-, SB-, FB-, A-, L-, DOC-) in the entry content
Create bidirectional links: Update
```
crossReferences
```
in
```
context/index.json
```
Update topic indexes: Add entry keywords to
```
topicIndex
```
,
```
productIndex
```
, and
```
phaseIndex
```
as appropriate
Log what was linked: Include cross-reference summary in the save confirmation

See

rules/context-graph.md

for the full cross-reference specification.