Przejdź do treści
Biblioteka skilli
AI i agenci · Automatyzacja

session-memory

/session-memory

Klasyfikuje sygnały z sesji według warstwowej taksonomii i zapisuje potwierdzone elementy w pamięci przy zakończeniu pracy.

Repozytorium
pensyve ↗
Licencja
Apache-2.0
Ostatnia zmiana repozytorium
04.10.2026

Ocena i bezpieczeństwo

90/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść25 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 04.10.2026, bez odwołań do wycofanych modeli i API.

Licencja

Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.

Co robi i kiedy

Klasyfikuje sygnały z sesji według warstwowej taksonomii i zapisuje potwierdzone elementy w pamięci przy zakończeniu pracy.

Oryginalny opis (EN)

End-of-session memory capture -- classifies session signals using a tiered taxonomy and stores confirmed items via Pensyve. Use when ending a work session or when the user wants to capture what was learned.

Jak działa

  1. Instructions
  2. Constraints
  3. Error Handling
Sekcje z nagłówków SKILL.md, w oryginale · 134 linii, 909 słów
Podgląd SKILL.md
---
name: session-memory
description: "End-of-session memory capture -- classifies session signals using a tiered taxonomy and stores confirmed items via Pensyve. Use when ending a work session or when the user wants to capture what was learned."
---

# Session Memory Capture

Analyze the current session using the tiered capture taxonomy, then store confirmed items in Pensyve memory.

## Instructions

When this skill is invoked (typically at the end of a coding session), follow these steps:

### Step 1: Analyze the Session Using Tiered Classification

Review the current session conversation and classify memorable content into two tiers:

**Tier 1 -- High-confidence candidates (confidence >= 0.9):**

These are high-signal items that should usually be offered for storage. User
confirmation is still mandatory before storing any item:

- User explicitly states a decision ("let's use X", "we decided Y", "we chose Z")
- User corrects agent behavior ("don't do X", "stop doing Y", "no, not that")
- User states a project constraint ("we can't use X because Y", "must not use Z")
- Technology migrations ("switching to X", "migrating from Y to Z")

**Tier 2 -- Review candidates (confidence 0.7-0.89):**

These are medium-signal items that benefit from user confirmation:

- Debugging session reveals root cause ("the bug was caused by...", "root cause is...")
- Approach tried and abandoned with reason ("tried X but it failed because...")
- Performance finding with measurable result ("query time dropped from 2s to 50ms")
- Cross-component dependency discovered ("X depends on Y", "blocked by Z")
- Workaround for framework/tool limitation ("workaround: use X instead of Y")

**Discard -- Never store:**

- Simple typo or formatting fixes
- Routine lint fixes, import sorting, boilerplate
- Standard file edits with no architectural significance
- Repeated application of already-known patterns
- Very short interactions that are clearly routine
- Content that is already stored in Pensyve (check via `pensyve_recall` with targeted queries)

### Step 2: Sanitize and Deduplicate Candidates

Before sending a candidate to `pensyve_recall` or presenting it to the user:

- remove anything that looks like an API key, token, password, or credential;
- exclude the candidate entirely and warn the user if removing sensitive data
  would destroy its meaning;
- truncate individual facts to 512 characters maximum; and
- summarize long code blocks rather than including them verbatim.

Then run `pensyve_recall` with a query matching the sanitized candidate fact.
If a highly similar memory already exists (score > 0.85), omit the candidate
and inform the user that it was skipped as a duplicate.

### Step 3: Present Candidates for Confirmation

Present the candidate memories to the user in a structured format, grouped by tier:

> **Session Memory Candidates**
>
> **Tier 1 -- High confidence (confirmation required):**
>
> 1. `auth-service`: Chose RS256 over HS256 for JWT signing to support key rotation (0.95)
> 2. `api-design`: POST endpoints return 201 with the created resource, not 200 (0.9)
>
> **Tier 2 -- For review:**
>
> 3. `database`: Migration script fails silently when Python < 3.11 -- added version check (0.8)
> 4. `testing`: Integration tests that touch the filesystem need tmpdir cleanup (0.7)
>
> Which items should I store? (e.g., "all", "1,3", "none")

### Step 4: Store Confirmed Items

For each confirmed item, decide the storage type based on its tier:

**Tier 1 items -> Semantic (durable facts).** Call `pensyve_remember` with:

- `entity`: The inferred entity name (lowercase, hyphenated)
- `fact`: `"[auto-capture/user/residual/tier-1] <fact text>"`
- `confidence`: 0.9-0.95 (based on classification)

Use for: architecture decisions, technology choices, user preferences, project constraints.

**Tier 2 items -> Episodic (observations).** Call `pensyve_observe` with:

- `episode_id`: From the session state (if episode tracking is active)
- `content`: `"[auto-capture/user/residual/tier-2] <observation text>"`
- `source_entity`: `"antigravity-cli"`
- `about_entity`: The inferred entity name (lowercase, hyphenated)
- `content_type`: `"text"` for decisions/patterns, `"code"` for code-related outcomes

Use for: root causes, failed approaches, performance findings, dependency discoveries.

If no episode is active, fall back to `pensyve_remember` with confidence 0.7-0.8.

When in doubt, prefer `pensyve_observe` -- the consolidation engine promotes recurring patterns to semantic facts automatically.

### Step 5: Close an Active Episode

If a working episode remains active, derive its outcome from the actual session
result, call `pensyve_episode_end`, and clear the working ID. Use `"success"`
only when the requested work completed, `"failure"` when it produced no usable
result, and `"partial"` when required work remains. Do not double-close an
episode already closed by another workflow.

### Step 6: Report Results

After storing, summarize what was saved:

> **Stored 3 memories:**
>
> - `auth-service`: Chose RS256 over HS256 for JWT signing (confidence: 0.9)
> - `database`: Migration script fails silently when Python < 3.11 (confidence: 0.8)
> - `testing`: Integration tests need tmpdir cleanup (confidence: 0.7)

## Constraints

- **NEVER auto-store.** Every candidate MUST be presented to the user for confirmation before calling `pensyve_remember` or `pensyve_observe`. This is a hard requirement.
- **Never read or write `.claude/` memory files.** All memory operations go through the Pensyve MCP tools exclusively.
- Entity names MUST be lowercase and hyphenated.
- Do not store secrets, API keys, passwords, or credentials. Warn the user if a candidate appears to contain sensitive data.
- If the session has no significant content worth remembering, say so clearly rather than forcing low-quality memories.
- When in doubt about whether something is significant, include it as a candidate and let the user decide.

## Error Handling

- If `pensyve_remember` fails, display the error and continue with remaining items.
- If `pensyve_observe` fails, display the error and continue with remaining items.
- If `pensyve_recall` (duplicate check) fails, proceed with storage but note that duplicate checking was skipped.
- If the MCP server is not connected, inform the user and suggest checking their MCP server configuration.

Wersje

RepozytoriumWynikLicencja
Ta strona · major7apps/pensyve90Apache-2.0
oaustegard/claude-skills100MIT
yo-steven/skills-exploration-20260522100Apache-2.0
major7apps/pensyve90Apache-2.0
major7apps/pensyve90Apache-2.0

Migawka bazy z 13.11.2026 · wersja paczki 894fdd18b055 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 90pensyve
    Przywołuje i zapisuje pamięć projektu w Pensyve oraz sprawdza jej stan, gdy użytkownik o to poprosi.
  • 90keep-memory-current
    Odczytuje na początku pracy zapisy pamięci projektu i zapisuje to, czego się nauczono, przed zakończeniem sesji.
  • 100mention-workflow
    Obsługuje przepływ z wzmianką @pensyve w Codex, dający dostęp do pamięci Pensyve w stylu wzmianek.
  • 100memory-capture
    Ujednolica zapisywanie pamięci z routingiem: kompaktowanie sesji lub szybkie wskazówki, wywoływane poleceniami o zapisie i kompresji sesji.
  • 90keep-memory-current
    Wczytuje pamięć projektu na początku sesji i zapisuje wnioski przed zakończeniem, także po trudnym debugowaniu.
  • 100session-capture
    Tworzy znormalizowany zapis sesji z celem, wykonaną pracą, decyzjami, blokadami i wynikami weryfikacji.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.