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 ChrisLamDev/hermes-core-skills --skill context-compaction-verification-and-recoverygit clone --depth 1 https://github.com/ChrisLamDev/hermes-core-skillsWrote 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/chrislamdev/hermes-core-skills/context-compaction-verification-and-recovery)<a href="https://agentmods.dev/skills/chrislamdev/hermes-core-skills/context-compaction-verification-and-recovery"><img src="https://agentmods.dev/badge/skills/chrislamdev/hermes-core-skills/context-compaction-verification-and-recovery/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/chrislamdev/hermes-core-skills/context-compaction-verification-and-recovery"><img src="https://agentmods.dev/badge/skills/chrislamdev/hermes-core-skills/context-compaction-verification-and-recovery.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00021 | $0.02695 |
| Opus 5 | $0.00010 | $0.01347 |
| Sonnet 5 | $0.00004 | $0.00539 |
| Haiku 4.5 | $0.00002 | $0.00269 |
Grade A, and why
context-compaction-verification-and-recovery 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 11d 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 — 225 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Context Compaction Verification & Recovery Protocol
When to Use
Use this protocol when:
- You've issued multiple
patch,write_file, orcreate_filecalls in a session - Context compaction has occurred (handoff summary mentions "completed" work)
- User reports "偏翼唔到出嚟" (page doesn't render / doesn't work)
- You suspect some tool calls were "spoken but not executed" — written in responses but not actually made
Core Insight
Tool calls that exist only in your reply text but weren't actually invoked will NOT have modified any files. This is the most common failure mode during complex multi-step sessions across context compactions.
Pattern: AI describes what it will do in markdown text blocks, the user says "繼續", and in the next turn the AI treats that description as already done without having made the actual tool calls.
Warning Signs of Compaction-Induced Drift
These patterns strongly suggest patches never landed:
- Deferred tool calls — you say "first I'll X, then Y" in prose, then move on without actually calling the tools
- "As mentioned above" syndrome — you reference code that was typed in your response but never written to disk
- Excessive prose descriptions of code — if you're describing implementation in detail rather than doing it, check if tools were actually called
- User says it doesn't render after you think you've finished — this is the ultimate signal
Prevention: The "Did the tool fire?" Check
Before any "繼續" transition, mentally scan: for every file I claim to have modified, did I actually see a patch() or write_file() success response in my tool output? If you only remember typing it in prose, it didn't happen.
Step 1: Audit — File Existence Check
Before any repairs, run a comprehensive audit:
# Find all NEW files that should exist
# Check each one with ls or search_files
Key files to verify:
- New utility files (
utils/*.js) - New page directories (
pages/*/) - New component files
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.
- 11d ago First seen · 225 lines · 21 tokens per session scan A f7274e494a59
context-compaction-verification-and-recovery is a skill published in the GitHub repository ChrisLamDev/hermes-core-skills (9 stars, last pushed 2mo ago), licensed MIT. It adds 21 tokens to every session and 2,695 once invoked, about $0.0001 per session on Opus 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 2026-08-31.
Other skills, from other repositories
unicli-repair
Evidence-driven repair workflow for a broken Uni-CLI adapter. Trigger on a failed unicli envelope, a quarantined adapter, or an explicit adapter-repair request. Classifies non-source failures, edits only the reported adapter path, and uses the original command as a bounded oracle.
autonomous-run
Prepare, start, inspect, resume, or stop a finite local overnight coding run after a human has accepted a Wayfinder terminal spec; coordinates a declared Claude/Codex maker and independent checker without pushing, merging, or writing to external systems.
research-archival
Scrape AI research URLs, archive with frontmatter, create GitHub Issues with identity verification.
authentication-patterns
OAuth 2.0, JWT, SSO, MFA, NextAuth/Clerk/Supabase Auth implementation patterns.
case-interview-practice
Interactive consulting case interview practice with structured frameworks, feedback mechanisms, and progressive difficulty. Use when preparing for management consulting interviews, case competitions, or business problem-solving exercises.
i18n-localization
Internationalization and localization for global applications. Use when adding multi-language support, handling regional formats, or preparing apps for global markets.