Generate falsifiable hypotheses from session signals and rank them by evidence.
npx skills add https://github.com/mifunedev/openharness --skill retro-mifunedev
Please help me install this Agent Skill. Skill: retro Source: https://github.com/mifunedev/openharness/tree/main/.claude/skills/retro Command: npx skills add https://github.com/mifunedev/openharness --skill retro-mifunedev
This skill helps teams and agents turn ongoing conversation signals into testable hypotheses, capture evidence, and decide what should be promoted into durable memory or identity constraints.
Invoke /retro to perform a session reflection and memory gating pass.
High-intent search queries and answers about installing and using this skill.
Evidence-based session reflection analyzes active conversation signals to generate falsifiable hypotheses, scoring them by supporting and refuting evidence before promoting durable insights into memory files.
You convert conversation signals into testable hypotheses by applying a six-subsystem lens to the active session, ranking generated hypotheses by evidence, and promoting only well-supported insights to durable memory.
Session reflection requires a SKILL.md frontmatter containing name and description fields, and writes promoted memory gating results directly to memory/MEMORY.md and context/IDENTITY.md files.
You trigger a session reflection and memory gating pass by invoking the /retro command, which analyzes the current conversation and writes well-supported hypotheses to memory and identity files.
The best way to decide which insights promote to durable memory is scoring falsifiable hypotheses by supporting and refuting evidence, then gating only medium-or-higher confidence results into memory files.
You should not use automated memory gating when your session lacks decision-heavy signals or testable outcomes, as the reflection process requires sufficient evidence to generate and rank falsifiable hypotheses.
aRustyDevGenerate testable hypotheses from observations and design experiments.
maystudiosStore and retrieve project learnings and decisions using GitHub Issues.
JKHeadleyRestore persistent agent identity from .instar files after session restarts.
linenoizeCapture and retrieve cross-session project memory via semantic memory MCP operations.
