Query past interactions — agent conversations, gateway sessions, and skill invocations — to understand what was discussed, who contributed, and what conclusions were reached.

V2V Phase

Cross-phase — This skill surfaces conversation history relevant to any phase of work.

Prerequisites: Interaction log populated (interactions logged via the interaction logging system) Outputs used by: Any agent or skill needing prior conversation context

Purpose

Before starting work in an area that may have been discussed before, query the interaction log to surface:

•What was previously discussed on this topic
•Which agents contributed perspectives
•What conclusions or outcomes were reached
•What open questions remain

This complements /context-recall (which searches decisions, bets, and learnings) by searching conversation history — the request/response pairs that led to those decisions.

When to Use

Invoke /interaction-recall [topic] when:

•Starting a new session and want to pick up where you left off
•Checking what agents previously said about a topic
•Auditing how a conclusion was reached
•Understanding the progression of thinking on a topic
•Finding which agents have weighed in on an area

Query Syntax

code

/interaction-recall pricing                      → All pricing interactions
/interaction-recall pricing agent:pm             → Only PM's pricing interactions
/interaction-recall pricing date:2026-01         → January 2026 only
/interaction-recall pricing product:AXIA         → Product-filtered
/interaction-recall pricing type:gateway         → Only gateway sessions
/interaction-recall agent:pm                     → All PM interactions (any topic)
/interaction-recall date:2026-01-27              → All interactions on a specific date

Filters

Filter	Format	Example
`agent:`	Agent key	`agent:pm`, `agent:vp-product`, `agent:plt`
`date:`	YYYY, YYYY-MM, or YYYY-MM-DD	`date:2026`, `date:2026-01`, `date:2026-01-27`
`product:`	Product name	`product:AXIA`
`type:`	Interaction type	`type:agent`, `type:gateway`, `type:skill`

Multiple filters can be combined: /interaction-recall pricing agent:pm product:AXIA

Process

1. Parse Query

Extract:

•Topic keywords: The main search terms
•Filters: agent, date, product, type (optional)

2. Search JSON Index

Read context/interactions/index.json and search:

•Topic search: Match keywords against topicIndex keys and entry topics arrays
•Request/outcome search: Match keywords against requestSummary and outcomeSummary fields
•Apply filters: Filter results by agent, date, product, type as specified

3. Rank Results

Order by relevance:

•Exact topic match in topicIndex
•Keyword match in requestSummary or outcomeSummary
•Partial topic match
•Most recent first (within same relevance tier)

4. Retrieve Full Content

For the top 5 matches:

•Read the daily log file at entries[].path
•Find the specific entry by ID within the file
•Extract full request and response content

5. Present Results

markdown

## Interaction History: [Topic]

*Found [N] interactions matching "[query]"*
*Filters: [applied filters]*

---

### IX-2026-00042 | {emoji} @pm | 2026-01-27 14:23
**Topics**: prd, authentication, requirements-gaps
**Product**: AXIA

**Request**: Review authentication PRD for requirements gaps

**Outcome**: Found 3 gaps: SSO edge cases, session timeout, MFA fallback

**Related**: DR-2026-003, DOC-2026-015

<details>
<summary>Full conversation</summary>

[Full request and response text]

</details>

---

### IX-2026-00038 | {emoji} @vp-product | 2026-01-25 10:15
...

---

### Related Context

These skills may surface additional information:
- `/context-recall [topic]` — Decisions, bets, learnings on this topic
- `/feedback-recall [topic]` — Customer feedback on this topic

*Showing top 5 of [N] matches. Narrow with filters: `agent:`, `date:`, `product:`, `type:`*

6. Handle No Results

If no interactions match:

markdown

## Interaction History: [Topic]

No interactions found matching "[query]".

**Suggestions**:
- Try broader keywords
- Remove filters to widen search
- Check `/context-recall [topic]` for decisions and documents
- Check `/feedback-recall [topic]` for customer feedback

Instructions

•Accept topic/keyword from user (at least one keyword or filter required)
•Parse optional filters: agent:, date:, product:, type:
•Read context/interactions/index.json
•Search entries using topic matching + filters
•Rank results by relevance, then recency
•For top 5 matches, read full content from daily log files
•Present results with summaries and expandable full content
•Suggest related context skills (/context-recall, /feedback-recall)
•If no matches, provide helpful suggestions

Example Usage

code

User: /interaction-recall pricing

Agent: Searches context/interactions/index.json for topic "pricing"
       Returns matching interactions with summaries and full content
       Suggests /context-recall pricing for related decisions

code

User: /interaction-recall agent:pm date:2026-01

Agent: Returns all PM interactions from January 2026
       Ordered by date, most recent first

code

User: /interaction-recall enterprise product:AXIA type:gateway

Agent: Returns gateway sessions about enterprise topics for AXIA
       Shows all agent perspectives from those sessions