Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add vasilyu1983/AI-Agents-public --skill agents-skills-feedback-loopgit clone --depth 1 https://github.com/vasilyu1983/AI-Agents-publicWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/vasilyu1983/ai-agents-public/agents-skills-feedback-loop)<a href="https://agentmods.dev/skills/vasilyu1983/ai-agents-public/agents-skills-feedback-loop"><img src="https://agentmods.dev/badge/skills/vasilyu1983/ai-agents-public/agents-skills-feedback-loop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/vasilyu1983/ai-agents-public/agents-skills-feedback-loop"><img src="https://agentmods.dev/badge/skills/vasilyu1983/ai-agents-public/agents-skills-feedback-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Memory Poisoning · line 19 Skill injects content designed to persist in agent memory or context across interactions. Persistent injection can alter agent behavior long after the initial interaction.Fix: Do not allow untrusted input to persist in agent memory or context. Validate all content before storing and implement memory isolation between sessions.
- medium Rogue Agent · line 38 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00035 | $0.04606 |
| Opus 5.5 | $0.00014 | $0.01842 |
| Sonnet 5.5 | $0.00007 | $0.00921 |
| Haiku 4.5 | $0.00003 | $0.00461 |
Grade A, and why
agents-skills-feedback-loop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 17d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 241 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Skills — Feedback Loop
Use this skill to wire a learnings loop into another skill so it gets better with use, without rewriting SKILL.md automatically.
The loop has four moving parts:
learnings.md— raw, append-only, committed. Shared working memory across machines; created on first append viaappend_learning.py, not seeded empty.learnings.consolidated.md— pruned, dated, committed. Portfolio-grade institutional memory; seeded at wiring time.learnings.local.md— machine-specific notes, gitignored. Use for one-operator-on-one-machine context that should not propagate.scripts/append_learning.py+scripts/consolidate.py— keep raw entries well-shaped and redacted, and surface recurring ones for human-reviewed consolidation.
Pattern provenance and the distinction between this reviewed loop and automatic skill rewriting are in references/evidence-base.md; load it before attributing the design to an external source.
Quick Reference
| Task | Read or Run | Outcome |
|---|---|---|
| Wire a skill to use the loop | references/wiring-protocol.md |
Adds a 4-line addendum to that skill's SKILL.md, seeds files |
| Format a new learning entry | references/learnings-format.md |
Atomic, dated, 5-section schema that survives pruning |
| Promote raw → consolidated | python3 scripts/consolidate.py <skill-dir> |
Dry run by default: proposes dedup, age-out and recurring patterns for human review; --apply only prints the edits to make, because promotion stays a human gate |
| Append a learning safely | python3 scripts/append_learning.py <skill-dir> --section <name> --text "..." |
Validates shape, dates, refuses to grow past the 150-entry cap |
| Find routing lessons that need review | python3 scripts/consolidate.py --queue [--days 14] |
Exit 1 for aged open lessons, failed results, or fixes awaiting verification; invalid dates exit 2 |
| Audit drift across skills | references/audit-checklist.md |
Find stale loops, missing consolidations, oversized files |
| Capture a cross-cutting behavior with no obvious skill home | references/instinct-pattern.md |
Atomic, confidence-scored complement to the per-skill loop above |
What ships with it
22 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 372 B
- assets/learnings_capture.py 12 KB runs code
- assets/learnings.template.md 762 B
- data/sources.json 5.8 KB
- learnings.consolidated.md 775 B
- learnings.md 558 B
- references/audit-checklist.md 4.9 KB
- references/closed-loop-capture.md 8.2 KB
- references/consolidation-protocol.md 3.1 KB
- references/evidence-base.md 8.0 KB
- references/instinct-pattern.md 5.6 KB
- references/learnings-format.md 2.5 KB
- references/wiring-protocol.md 3.3 KB
- scripts/append_learning.py 8.3 KB runs code
- scripts/bulk_wire.py 4.2 KB runs code
- scripts/consolidate.py 12 KB runs code
- scripts/install_capture_hook.py 5.2 KB runs code
- scripts/test_append_learning.py 12 KB runs code
- scripts/test_bulk_wire.py 1.7 KB runs code
- scripts/test_capture_hook.py 4.2 KB runs code
- scripts/test_consolidate_queue.py 6.4 KB runs code
- scripts/test_consolidate.py 3.3 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 17d ago Changed e49523996f8c
- 24d ago Changed · -89 lines 423e5d603f73
- 1mo ago Changed · +2 lines b0d3504116d8
- 2mo ago First seen · 328 lines · 35 tokens per session scan A 7fa91f747985
agents-skills-feedback-loop is a skill published in the GitHub repository vasilyu1983/AI-Agents-public (101 stars, last pushed 14d ago), licensed MIT. It adds 35 tokens to every session and 4,606 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
principle-guard-the-context-window
Apply when context is filling up: large outputs, long files, repeated reads, fan-out planning. Route bulk to subagents; keep summaries in the main thread, not raw payloads.
continuum-memory
Configure and use Continuum's two-tier memory system — mem0+Qdrant/Milvus for long-term facts, Redis for short-term sessions, with multi-tenant scopes (USER / AGENT / SHARED / RUN / CONVERSATION). Invoke when the user asks about "remember", "user preferences", "long-term memory", "vector search over memories"…
dual-brain
Dual-brain review for ambiguous designs, risky refactors, and recurring project decisions. Combine macro-level design review with micro-level code verification and relevant project memory.
session-handoff
Context transfer between AI sessions. Trigger when the user wants to save the context, resume a task, or hand off the work to another session.
memory
How to save what you learn about people, promises and investigations, and what never to save.
context
Context lifecycle as thin seams over native primitives (/foundry:context ). WRAP of the retired context-lifecycle skills — additive over native /compact, --resume, /rewind, and the context sensor; never reinvents transcript replay. Trigger to snapshot/resume a session's distilled arc-state or check context budget.