Skills self-improving-agent
๐Ÿ“ฆ

self-improving-agent

Content revision r2 Medium Risk โš™๏ธ External commands๐ŸŒ Network access๐Ÿ“ Filesystem access

Improve Skills From Validated Feedback

Repeated workflow lessons can be lost or applied without enough evidence. This skill captures structured experiences and creates traceable proposals before durable guidance changes.

Supports: Claude Codex Code(CC)
๐Ÿ“Š 70 Adequate

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request
Review the Skillstore skill "self-improving-agent" from https://skillstore.io/skills/charon-fan-self-improving-agent.md and its manifest at https://skillstore.io/api/skills/charon-fan-self-improving-agent/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

Your Agent should still show its plan and request any confirmation required by the security policy.

Test it

Using "self-improving-agent". A debugging workflow found an empty refresh callback in three separate incidents.

Expected outcome:

  • Pattern: verify callback implementations before release.
  • Evidence: three documented incidents with the same root cause.
  • Next step: propose a debugger checklist update and validate it with a focused test.

Using "self-improving-agent". A user reports that a generated deployment guide was incomplete.

Expected outcome:

  • Episode: deployment guide omitted rollback verification.
  • Confidence: low until the guide is reproduced and reviewed.
  • Proposal: add a rollback checklist only after validation.

Security Audit

Medium Risk
v8 โ€ข 7/23/2026 Open versioned report

Static findings are predominantly false positives caused by Markdown code fences, inline backticks, research links, and documentation of local paths. The bundled hook scripts only log event metadata and output lengths; they do not perform network requests, command interpolation, or memory writes. The documented automatic memory workflow still needs privacy controls because it can retain context from many skill events.

10
Files scanned
1,091
Lines analyzed
0
Review items
0
False positives ignored

Confirmed security concerns (1)

Medium
Automatic cross-skill memory collection may retain sensitive task context
The skill describes automatic capture of experiences from all skill events and memory artifacts, but does not define data minimization, retention, or consent controls. Users may record task details or tool-related context in a persistent local memory store.
The documentation explicitly states universal event learning and identifies persistent semantic, episodic, and working memory locations. It does not specify a retention or sensitive-data exclusion policy.
Audited by: codex View Audit History โ†’
Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report
Security Assessment

Copy report link

https://skillstore.io/skills/charon-fan-self-improving-agent/audits/8?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/charon-fan-self-improving-agent/security.svg)](https://skillstore.io/skills/charon-fan-self-improving-agent?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/charon-fan-self-improving-agent?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/charon-fan-self-improving-agent/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/charon-fan-self-improving-agent.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>
Academic citations (APA ยท BibTeX ยท CFF)

APA citation

charon-fan. (2026). self-improving-agent security audit report (audit version 8) [Author version unspecified]. Skillstore. https://skillstore.io/skills/charon-fan-self-improving-agent/audits/8

BibTeX citation

@techreport{charon-fan-charon-fan-self-improving-agent-2026, author = {charon-fan}, title = {self-improving-agent security audit report (audit version 8)}, institution = {Skillstore}, year = {2026}, number = {8}, url = {https://skillstore.io/skills/charon-fan-self-improving-agent/audits/8}, note = {Author version unspecified} }

CITATION.cff

cff-version: 1.2.0 message: "If you use this Skill, cite its author and this versioned security audit report." title: "self-improving-agent security audit report (audit version 8)" version: "unspecified" type: report authors: - name: "charon-fan" date-released: "2026-07-23" url: "https://skillstore.io/skills/charon-fan-self-improving-agent/audits/8" identifiers: - type: other value: "skillstore:charon-fan-self-improving-agent:audit:8" description: "Skillstore immutable audit report identifier"

Skillstore Score

Why this score Evidence Confidence: High
55
Architecture
85
Maintainability
87
Content
75
Community
83
Spec Compliance

What You Can Build

Maintain team playbooks

Turn recurring support and review lessons into traceable proposals for shared skill guidance.

Learn from debugging sessions

Record verified root causes and promote repeated debugging patterns after review.

Review agent workflows

Collect feedback signals and validate proposed changes before updating agent instructions.

Try These Prompts

Capture one lesson
Review this completed workflow. Extract one verified lesson, its evidence, and a safe proposal for future use.
Summarize feedback
Summarize these user feedback notes into episodic memories. Separate facts from assumptions and assign confidence.
Compare recurring patterns
Compare these episodes. Identify patterns that recur, explain the supporting evidence, and propose only low-risk guidance updates.
Prepare an approval-ready proposal
Create a proposal to improve this skill. Include affected files, evidence, validation steps, rollback notes, and ask for approval before durable edits.

Best Practices

  • Capture facts separately from interpretations and record supporting evidence.
  • Promote durable changes only after validation or explicit user approval.
  • Review stored memory regularly and remove sensitive or outdated artifacts.

Avoid

  • Do not generalize a single incident into a permanent rule.
  • Do not overwrite skills, repository rules, or runtime code without approval and validation.
  • Do not store credentials, private data, or unnecessary tool output in memory artifacts.

Frequently Asked Questions

What does this skill improve?
It helps convert verified workflow lessons into memory artifacts and improvement proposals.
Does it edit skills automatically?
It should capture proposals first. Durable guidance changes require validation or explicit approval.
What information can it store?
It can store structured lessons, outcomes, feedback, and proposed improvements. Avoid sensitive data.
Are hooks required?
No. Hooks are optional and depend on the host runtime and local configuration.
How are patterns promoted?
Patterns should recur or be validated by focused tests or reviews before promotion.
Can I remove stored learning data?
Yes. Review and delete local memory artifacts according to your retention policy.

Developer Details

Author

charon-fan

License

MIT

Skillstore revision

r2

Version notice

The author did not declare a version.

Ref

c43861a65bb95efcae259cd161c9d6f4dc7eec6f

Maintenance freshness

7/24/2026

Usage

122 downloads ยท 234 views

File structure

๐Ÿ“ hooks/

๐Ÿ“„ post-bash.sh

๐Ÿ“„ pre-tool.sh

๐Ÿ“„ session-end.sh

๐Ÿ“ memory/

๐Ÿ“„ semantic-patterns.json

๐Ÿ“„ README.md

๐Ÿ“ references/

๐Ÿ“„ appendix.md

๐Ÿ“„ SKILL.md

๐Ÿ“ templates/

๐Ÿ“„ correction-template.md

๐Ÿ“„ pattern-template.md

๐Ÿ“„ validation-template.md

View all