Borrowing it
Nothing to install: this file belongs to rodspeed/epistemic-memory. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/rodspeed/epistemic-memory/master/.claude/skills/observe/SKILL.mdgit clone --depth 1 https://github.com/rodspeed/epistemic-memoryWrote 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/rodspeed/epistemic-memory/observe)<a href="https://agentmods.dev/skills/rodspeed/epistemic-memory/observe"><img src="https://agentmods.dev/badge/skills/rodspeed/epistemic-memory/observe.svg" alt="Measured on agentmods" 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 | $0.00000 | $0.01502 |
| Opus 5 | $0.00000 | $0.00751 |
| Sonnet 5 | $0.00000 | $0.00300 |
| Haiku 4.5 | $0.00000 | $0.00150 |
Grade A, and why
observe 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 8d 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 — 111 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Observe — Passive Behavioral Evidence Collector
Automatically collects evidence about the user from how they work — not what they say about themselves, but what their choices reveal. Runs silently at conversation end alongside /harvest. Writes to an evidence log, not directly to the portrait.
This skill is never invoked by the user. It runs automatically as a step in your end-of-session harvest — the two are a single atomic operation. Can also be invoked standalone for manual catch-up. No output is shown unless the observation is worth surfacing (see Step 4).
Trigger Condition
Runs at conversation end when the session was substantive (same gate as harvest) AND involved at least one of:
- The user made a judgment call where reasonable people would disagree
- The user pushed back on a suggestion, revealing a priority or value
- The user's behavior contradicted something in the existing profile
- The user revealed something personal unprompted (not in response to
/mirror interview) - The user's process (how they worked, what they skipped, what they lingered on) showed something not captured in the profile
If none of these occurred — a purely mechanical session — write nothing. The absence of signal is not signal.
Instructions
Step 1: Load Current Profile (silent)
Read these files to know what's already captured:
memory/user/— all profile files (identity, drives, creative life, etc.)memory/user/tensions.mdmemory/user/observations.md(the evidence log — check for duplicates)
Paths are relative to your project's memory root.
Step 2: Scan the Conversation (silent)
Review the full conversation. You're looking for behavioral evidence, not self-report. The distinction matters:
| Type | Example | What to record |
|---|---|---|
| Judgment under ambiguity | Chose to foreground one approach over another despite the alternative having stronger surface-level support | "Prioritized X over Y when structuring an argument — reveals weighting of [value]" |
| Pushback | Rejected a suggestion to add more caveats | "Resists hedging when the data supports a claim — prefers confident statement + honest limitation section over distributed uncertainty" |
| Process signal | Spent 20 minutes on a sentence, moved on from a section in 30 seconds | "Allocates attention to [X], trusts instinct on [Y]" |
| Contradiction | Profile says X, but this session they did Y | Log as tension seed with both data points |
| Unprompted reveal | Mentioned something about family, past, feelings while working | Record the reveal and what it might connect to |
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.
- 8d ago First seen · 111 lines · 0 tokens per session scan A b130a21742ea
observe is a skill published in the GitHub repository rodspeed/epistemic-memory (20 stars, last pushed 5mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,502 tokens. 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
weekly-digests
Generate a serial week-by-week narrative digest of a project's full claude-mem timeline. Splits the timeline into per-ISO-week files, then runs one consecutive subagent per week — each receiving the prior week's carry-forward block — to produce one chapter per ISO week of data. Use when asked for "weekly digests"…
cloud-sync
Set up or check claude-mem cloud sync with cmem.ai Pro. Use when the user says "set up cloud sync", "sync my memories", "cmem pro", "cloud backup", "sync status", or wants their memory database backed up or synced to their cmem.ai account.
hivemind-memory
Global team and org memory powered by Activeloop. ALWAYS check BOTH built-in memory AND Hivemind memory when recalling information.
memory-audit-pattern-extraction
A method for investigating repeated mistakes by comparing related memories and checking whether an earlier reminder failed. It looks at where the reminder was stored, when it was created, and whether it was strong enough to prevent the mistake.
steer
Project-profile recalibration. Audits session calibration drift, presents per-cluster evidence for a user verdict, writes the updated project-profile rule, and inscribes a settled-direction clause.
recollect
Resolve vague recall into recognized context through AI-guided contextual scan and user-validated recognition — one session, or the line of work, topic, or settled concept spread across several.