Agent skill
reflect
End-of-session reflection protocol for llm-reproducibility. This skill should be used when the user invokes /reflect or asks to reflect on this session, do end-of-session reflections, or similar. It guides Claude through updating reflection documents in priority order, answering structured prompts, and maintaining research observation logs.
Install this agent skill to your Project
npx add-skill https://github.com/saross/llm-reproducibility/tree/main/.claude/skills/reflect
SKILL.md
End-of-Session Reflection
Update the project's reflection and observation documents following the established protocol. Work through the documents in priority order. If context is limited, prioritise the top of the list.
Important: Instance Boundary
Reflections are most valuable when written by the instance that did the session's work. If this invocation follows a compaction or continuation (i.e., the current instance is working from a conversation summary rather than direct experience), flag this explicitly in the reflection entries. Distinguish between genuine first-person observations and plausible reconstructions from summaries.
Protocol
Process all .md files in the reflections directory (docs/notes/reflections), sorted by
the priority field in their YAML frontmatter (lowest number = highest priority).
For each document: read it first to understand the current structure, entry numbering, and conventions, then append a new dated section continuing the established format.
Conditional documents
Some documents have scope: conditional in their frontmatter with a conditions field
describing when they should be updated. For these documents, evaluate whether the current
session meets the conditions. If not, explicitly state the assessment and skip.
Standards
- UK/Australian English throughout
- Concise but substantive — these are research documents
- Continue existing numbering sequences (do not restart)
- Include dated section headers matching the established format
- Update document footers/timestamps where they exist
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
research-assessor
Extracts and assesses research methodology, claims, and evidence from research papers in HASS disciplines. Evaluates transparency, replicability, and credibility through systematic extraction of research designs, methods, protocols, claims, and evidence using a six-pass iterative workflow.
research-assessor
Extracts and assesses research methodology, claims, evidence, and infrastructure from research papers in HASS disciplines. Evaluates transparency, reproducibility, and credibility through systematic extraction (eight-pass workflow, Pass 0-7) and credibility assessment (research approach classification, quality gating, and repliCATS Seven Signals evaluation adapted for HASS).
reproduction-assessor
Reproduces computational research papers through systematic Docker-based execution, quantitative verification, and adversarial review. This skill should be used when the user requests reproduction of a paper, verification of computational results, or assessment of reproducibility infrastructure. Handles R analyses with deterministic or stochastic outputs via a 4-session workflow (R-Plan, R-A, R-B, R-C). R-based papers only in v1.0; extension points for Python/Julia documented.
obsidian-vault
Search, create, and manage notes in the Obsidian vault with wikilinks and index notes. Use when user wants to find, create, or organize notes in Obsidian.
edit-article
Edit and improve articles by restructuring sections, improving clarity, and tightening prose. Use when user wants to edit, revise, or improve an article draft.
handoff
Compact the current conversation into a handoff document for another agent to pick up.
Didn't find tool you were looking for?