reflexion
Cross-session learning through guided dialogue. Extracts session insights and integrates into persistent memory when session knowledge should be preserved. Alias: Reflexion.
55
44%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./reflexion/skills/reflexion/SKILL.mdQuality
Discovery
35%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
The description conveys a general sense of what the skill does (preserving session knowledge) but relies on abstract, system-level terminology rather than concrete actions or natural user language. It lacks an explicit 'Use when...' clause with specific trigger scenarios, and the keywords used are unlikely to match how users would naturally request this functionality.
Suggestions
Add a 'Use when...' clause with natural trigger terms like 'remember this for next time', 'save what we learned', 'don't forget this', or 'carry over to future sessions'.
List specific concrete actions such as 'summarizes key decisions, captures learned preferences, updates memory files with new patterns, and recalls prior session context'.
Replace jargon like 'guided dialogue' and 'session insights' with user-facing language describing what gets remembered (e.g., 'preferences, decisions, project context, coding patterns').
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Names the domain ('cross-session learning', 'persistent memory') and some actions ('extracts session insights', 'integrates into persistent memory'), but the actions are somewhat abstract rather than listing multiple concrete operations. | 2 / 3 |
Completeness | The 'what' is partially addressed ('extracts session insights and integrates into persistent memory'), but the 'when' is only vaguely implied ('when session knowledge should be preserved') without explicit trigger guidance or a 'Use when...' clause with concrete scenarios. | 2 / 3 |
Trigger Term Quality | The terms used ('cross-session learning', 'guided dialogue', 'session insights', 'persistent memory') are technical/system-level jargon, not natural keywords a user would say. A user would more likely say 'remember this', 'save for next time', or 'learn from this session'. The alias 'Reflexion' is helpful only if users know that term. | 1 / 3 |
Distinctiveness Conflict Risk | The concept of 'persistent memory' and 'cross-session learning' provides some distinctiveness, but it could overlap with general memory/note-taking skills or knowledge management skills. The alias 'Reflexion' adds some uniqueness but the core description remains somewhat generic in the memory/learning space. | 2 / 3 |
Total | 7 / 12 Passed |
Implementation
54%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill has excellent workflow structure with clear 5-phase sequencing, explicit validation checkpoints, and well-organized progressive disclosure to external references. However, it suffers significantly from verbosity — the formal categorical notation, protocol integration details, and extensive inline decision trees consume many tokens that could be trimmed or moved to reference files. Actionability is moderate; while specific paths and tool calls are provided, much relies on custom abstractions (subagent types, Task syntax) without fully executable examples.
Suggestions
Move the formal notation block (Reflexion(S, M) → ...) entirely to references/formal-semantics.md — it adds no actionable value in the main skill file and consumes significant tokens.
Trim the Protocol Integration section to 1-2 lines each or move to a separate reference file; the cross-session enrichment pathway paragraph is especially verbose for the main skill body.
Condense the frontmatter update rules table and YAML example — the before/after merge example could be replaced with a single concise rule statement since Claude can infer YAML merge semantics.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is extremely verbose with extensive formal notation (categorical/type-theoretic), protocol integration details for other systems (Prothesis, Syneidesis), and detailed frontmatter merge rules that bloat the content significantly. The formal definition block and cross-session enrichment pathway explanation add substantial token cost with minimal actionable value. | 1 / 3 |
Actionability | The workflow phases provide concrete steps with specific file paths, JSON structures, and tool calls (TodoWrite, Task, AskUserQuestion), but much of the guidance is pseudocode-like rather than truly executable. The subagent invocations use a custom Task syntax that isn't standard, and critical details are deferred to external files (agents/*.md, references/). | 2 / 3 |
Workflow Clarity | The 5-phase workflow is clearly sequenced with explicit phase completion checkpoints (TodoWrite status updates at each transition), decision trees for user interaction (Q1-Q5), and a verification/cleanup phase. The decision branching is well-structured with clear paths for each user choice. | 3 / 3 |
Progressive Disclosure | Content is well-structured with a clear overview, phased workflow inline, and appropriate references to external files (agents/*.md, references/formal-semantics.md, references/memory-hierarchy.md, examples/worked-example.md). References are one level deep and clearly signaled in the Additional Resources section. | 3 / 3 |
Total | 9 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Repository
- jongwony/epistemic-protocols
- Commit
01bfef1
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.